├── README.md └── downloads ├── candc-1.00.tgz ├── candc-linux-1.00.tgz ├── candc-macosxu-1.00.tgz ├── grs2sd-1.00 ├── markedup_sd-1.00 ├── models-1.02.tgz ├── pos_bio-1.00.tgz ├── ptb_chunk-1.00.tgz ├── ptb_muc-1.00.tgz ├── ptb_pos-1.00.tgz └── super_bio-1.00.tgz /README.md: -------------------------------------------------------------------------------- 1 | _**Nota bene:**_ This project is entirely unsupported. 2 | I have not been able to get it to work, so creating an issue asking for guidance would be futile. 3 | If you have files to contribute, or build instructions / documentation that others might find useful, awesome! 4 | [Please submit a PR here](https://github.com/chbrown/candc/compare). 5 | 6 | # candc 7 | 8 | The canonical site, , is totally down. 9 | 10 | Luckily, the Wayback Machine has archived most of that site: 11 | 12 | * Latest capture of the [C&C tools main page](http://web.archive.org/web/20160318193242/http://svn.ask.it.usyd.edu.au/trac/candc) 13 | 14 | Unfortunately, James Curran put all the downloads behind a registration wall, so the Wayback Machine doesn't have any records of those. 15 | 16 | Fortunately, by the time I got around to creating this repository, Stephen Clark still hosted a copy of the downloads on his Cambridge website, ([latest capture on the Wayback Machine](https://web.archive.org/web/20180727140643/http://www.cl.cam.ac.uk/~sc609/candc-1.00.html)). 17 | The files available there were the 1.00 version, so not the latest version of everything, but better than nothing! 18 | 19 | ~That site and files are currently still live and accessible, but~ because I _knew_ that would happen, I downloaded and added them to this repository. The following section is taken verbatim from Stephen Clark's "C&C Parser Downloads" page. 20 | 21 | 22 | ## Version 1.0 9/06/07 23 | 24 | * source code: 25 | gzipped tar 26 | 27 | * precompiled binaries (do not contain Boxer): 28 | - Linux (static linked) gzipped tar (12MB) 29 | - Mac OS X 10.4 (universal) gzipped tar (12MB) 30 | 31 | * models trained on CCGbank 02-21 and MUC 7: 32 | gzipped tar (50MB) 33 | 34 | * other models: 35 | - POS model trained on Penn Treebank 3: 36 | gzipped tar (6MB) 37 | - chunk model trained on Penn Treebank 3: 38 | gzipped tar (4MB) 39 | - MUC 7 model (uses Penn Treebank 3 POS tags): 40 | gzipped tar (4MB) 41 | 42 | * models for biomedical parsing 43 | - POS model (v1.00, 1/12/09): 44 | gzipped tar (15MB) 45 | - supertagging model (v1.00, 1/12/09): 46 | gzipped tar (45MB) 47 | - markedup_sd file: 48 | markedup_sd 49 | - Post-processing script: grs2depbank (already current in parser 1.00) or grs2sd: 50 | grs2sd 51 | 52 | 53 | ## License 54 | 55 | I have no idea. Creative Commons should come up with something for abandonware. 56 | 57 | 58 | ## Other resurrections 59 | 60 | An [Inria researcher](https://github.com/valeriobasile) runs a live demo server at 61 | 62 | To repeat the needlessly heteronormative example from that page (_the patriarchy is strong with this one_): 63 | 64 | curl -d 'Every man loves a woman' 'http://gingerbeard.alwaysdata.net/candcapi/proxy.php/raw/pipeline?semantics=fol' 65 | 66 | > fol(1,not(some(A,and(n1man(A),not(some(B,some(C,and(r1patient(B,C),and(r1agent(B,A),and(v1love(B),n1woman(C))))))))))). 67 | -------------------------------------------------------------------------------- /downloads/candc-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/candc-1.00.tgz -------------------------------------------------------------------------------- /downloads/candc-linux-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/candc-linux-1.00.tgz -------------------------------------------------------------------------------- /downloads/candc-macosxu-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/candc-macosxu-1.00.tgz -------------------------------------------------------------------------------- /downloads/grs2sd-1.00: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | # Based on grs2depbank by James Curran 4 | # 5 | # Last updated 2 Sept 2008 LR 6 | 7 | import sys 8 | 9 | CONVERT = {'-LRB-': '(', '-RRB-': ')', 10 | '-LCB-': '{', '-RCB-': '}', 11 | '-LSB-': '[', '-RSB-': ']'} 12 | 13 | def punct_convert(s): 14 | for (old, new) in CONVERT.iteritems(): 15 | s = s.replace(old, new) 16 | return s 17 | 18 | def is_subset(list1,list2): 19 | subset = True 20 | for item in list1: 21 | if item not in list2: 22 | subset = False 23 | return subset 24 | 25 | PUNCT = set("( ) { } [ ] -- : ; , . ? !".split()) 26 | 27 | class Word: 28 | def __init__(self, token, index): 29 | self.index = index 30 | fields = token.split('|') 31 | self.token = fields[0] 32 | self.pos = fields[1] 33 | self.cat = fields[2] 34 | self.grs = [] 35 | self.sub = None 36 | 37 | def __cmp__(self, other): 38 | return cmp(self.index, other.index) 39 | 40 | def is_num(self): 41 | return self.pos == 'CD' 42 | 43 | def __repr__(self): 44 | if self.sub: 45 | return repr(self.sub) 46 | else: 47 | return '%s_%d' % (self.token, self.index) 48 | 49 | def convert_arg(arg, gr, words): 50 | if '_' in arg and arg != '_': 51 | word = words[arg] 52 | word.grs.append(gr) 53 | return word 54 | else: 55 | return arg 56 | 57 | def nopos(word): 58 | if type(word) == type(''): 59 | return word 60 | else: 61 | return word.token 62 | 63 | class GR: 64 | def __init__(self, line, words): 65 | self.line = line 66 | fields = line[1:-1].split() 67 | self.label = fields[0] 68 | self.args = [convert_arg(arg, self, words) for arg in fields[1:]] 69 | self.ignore = False 70 | def __repr__(self): 71 | return '(%s %s)' % (self.label, ' '.join(map(str, self.args))) 72 | 73 | def nopos(self): 74 | return '(%s %s)' % (self.label, ' '.join(map(nopos, self.args))) 75 | 76 | def replace(self, old, new): 77 | for i in xrange(len(self.args)): 78 | if self.args[i] == old: 79 | self.args[i] = new 80 | 81 | class Sentence: 82 | def __init__(self, grs, line): 83 | self.words = {} 84 | self.sentence = [] 85 | for (i, token) in enumerate(line.split()): 86 | word = Word(token, i) 87 | self.sentence.append(word) 88 | self.words[str(word)] = word 89 | 90 | self.grs = [GR(gr, self.words) for gr in grs] 91 | self.tokens = {} 92 | self.pos = {} 93 | self.cats = {} 94 | for w in self.sentence: 95 | self.tokens.setdefault(w.token, []).append(w) 96 | self.pos.setdefault(w.pos, []).append(w) 97 | self.cats.setdefault(w.cat, []).append(w) 98 | 99 | # NB subroutines from grs2depbank that aren't (currently) being used 100 | # are commented out 101 | def postprocess(self): 102 | if len(self.grs) == 0: 103 | return # some of these routines add GRs and we don't 104 | # want to do that if it was a parse failure 105 | self.rcmod_fix() # fixes a markedup bug 106 | self.question_aux_fix() # fixes a markedup bug 107 | self.det_to_poss() 108 | self.preconj() 109 | self.ecm_to_nsubj() 110 | self.add_funny_conj() 111 | self.fix_aux_conj() 112 | self.merge_conj_args() 113 | self.conj_to_appos() 114 | # self.remove_comma_conj() 115 | # self.add_passives() 116 | # self.ncmod_to_iobj() 117 | # self.fix_ncmod_num() 118 | self.num_and_number() 119 | # self.fix_question_poss() Don't need anymore, fixed in markedup 120 | self.in_vivo_vitro() 121 | self.switch_n_and_num() 122 | self.less_than() 123 | self.prep_to_appos() 124 | self.nn_and_amod() # REMOVE FOR BIOINFER 125 | # self.sfs_to_conj() 126 | # self.fix_ampersands() 127 | # self.remove_relpro_cmod() 128 | # self.cmod_to_ta() 129 | # self.ncmod_to_prt() 130 | # self.xmod_add_to() 131 | # self.fix_that_relpro() 132 | self.add_rel() 133 | # self.xmod_to_ta() 134 | # self.pct_ncmod_to_dobj() 135 | self.partmod_to_xcomp() 136 | self.only_dep() 137 | self.filter_punct() 138 | 139 | def rcmod_fix(self): 140 | """ fixes a bug in interpretation of markedup where 141 | a particular argument isn't output. Flagged with a special 142 | "rcmodfix" dependency label in markedup. 143 | """ 144 | grs = [gr for gr in self.grs if gr.label == 'rcmodfix'] 145 | for gr in grs: 146 | for gr2 in self.grs: 147 | if gr2.label == 'ref' and gr2.args[1] == gr.args[0]: 148 | gr.args[0] = gr2.args[0] 149 | gr.label = 'rcmod' 150 | 151 | def add_passives(self): 152 | for w in self.sentence: 153 | if w.cat.lstrip('(').startswith('S[pss]\NP') or \ 154 | (w.pos == 'VBN' and w.cat == 'N/N'): 155 | self.grs.append(GR('(passive %s)' % w, self.words)) 156 | 157 | def det_to_poss(self): 158 | """ converts det to poss when the token is a possessive 159 | """ 160 | poss = ['its', 'our', 'his', 'her', 'my', 'their', 'your'] 161 | for gr in self.grs: 162 | if gr.label == 'det' and gr.args[1].token in poss: 163 | gr.label = 'poss' 164 | 165 | def nn_and_amod(self): 166 | """ converts nmod to nn when pos=NN* and to amod when pos is 167 | anything else 168 | """ 169 | for gr in self.grs: 170 | if gr.label == 'nmod': 171 | if gr.args[1].pos.startswith('NN'): 172 | gr.label = 'nn' 173 | else: 174 | gr.label = 'amod' 175 | 176 | def fix_question_poss(self): 177 | """ fixes possessives in questions e.g. "what fruit's stone". 178 | they come out of the parser with "what" as an argument 179 | of the possessive rather than "fruit". 180 | specifically, changes: 181 | poss x y where y has pos=WDT 182 | to 183 | poss x z 184 | and 185 | possessive y w 186 | to 187 | possessive z w 188 | where in both cases z is the element such that there is 189 | a relation (det z y) 190 | """ 191 | for gr in self.grs: 192 | if gr.label == 'poss': 193 | if gr.args[1].pos == 'WDT': 194 | for gr2 in self.grs: 195 | if gr2.label == 'det' and gr2.args[1] == gr.args[1]: 196 | gr.args[1] = gr2.args[0] 197 | if gr.label == 'possessive': 198 | if gr.args[0].pos == 'WDT': 199 | for gr2 in self.grs: 200 | if gr2.label == 'det' and gr2.args[1] == gr.args[0]: 201 | gr.args[0] = gr2.args[0] 202 | 203 | def question_aux_fix(self): 204 | """ fixes a bug in interpretation of the markedup file, 205 | in which an aux is being produced in some relations where the 206 | main verb should actually be produced. 207 | Flagged with a special "nsubjfix" dependency label in markedup. 208 | """ 209 | grs = [gr for gr in self.grs if gr.label == 'nsubjfix'] 210 | for gr in grs: 211 | for gr2 in self.grs: 212 | if (gr2.label == 'aux' or gr2.label == 'cop') and gr2.args[1] == gr.args[0]: 213 | gr.args[0] = gr2.args[0] 214 | gr.label = 'nsubj' 215 | 216 | def add_funny_conj(self): 217 | for w in self.sentence: 218 | if w.cat == 'conj' and w.grs == []: 219 | prev = self.sentence[w.index - 1] 220 | gr = GR('(conj %s %s)' % (w, prev), self.words) 221 | self.grs.append(gr) 222 | # w.grs.append(gr) 223 | 224 | next = self.sentence[w.index + 1] 225 | gr = GR('(conj %s %s)' % (w, next), self.words) 226 | self.grs.append(gr) 227 | # w.grs.append(gr) 228 | 229 | def merge_conj_args(self): 230 | """ converts the rich parser output to more compact 231 | representation required by sd 232 | """ 233 | cc = {} # key = conjunction, value = list of its conjuncts 234 | # Add where the conjunction is a comma first 235 | for w in self.sentence: 236 | if w.token == ',': 237 | conj = [gr for gr in w.grs if gr.label == 'conj'] 238 | for gr in conj: 239 | cc.setdefault(str(w), []).append(gr.args[1]) 240 | gr.ignore = True 241 | 242 | # Remove subsets that happen with serial commas 243 | # e.g. for "a, b, c, and d" the parser will output conjunctions 244 | # consisting of c+d, b+c+d, a+b+c+d; we want to delete the 245 | # subsets and keep the superset 246 | for w,l in cc.iteritems(): 247 | for x,m in cc.iteritems(): 248 | if w != x and is_subset(l,m): 249 | cc[w].append('DELETE') # can't delete a dictionary entry while iterating through the dictionary - there should be cleaner ways to do this 250 | for w in cc.keys(): 251 | if 'DELETE' in cc[w]: 252 | del cc[w] 253 | 254 | # Add other conjunctions (conjunction is not a comma) 255 | # This is either the end of a list (x, y, and z / x, y and z), 256 | # a plain binary conjunction, or something like "is consistent 257 | # with, but does not prove". 258 | # At the end of a list, the subset check ensures that the conjunction 259 | # word gets used, but the original list is retained. 260 | # 261 | # Note that the parser already handles "x, y and z" by having the 262 | # comma conjoin all three items, otherwise this routine would 263 | # not work. 264 | addtocc = {} 265 | for w in self.sentence: 266 | if w.cat == 'conj': 267 | conj = [gr for gr in w.grs if gr.label == 'conj'] 268 | for gr in conj: 269 | gr.ignore = True 270 | conjuncts = [gr.args[1] for gr in w.grs if gr.label == 'conj'] 271 | if conjuncts: 272 | subset = False 273 | for x,l in cc.iteritems(): 274 | if is_subset(conjuncts,l): 275 | subset = True 276 | if 'DELETE' not in l: # if DELETE is there, the full list of conjuncts has already been added to addtocc with some conjunction word and we won't add another one (this usually happens when the parser has done something wrong with the coord anyway) (if we don't take this step, the list with DELETE in it gets into addtocc) (again there will be a cleaner way to do this) 277 | addtocc.setdefault(str(w), l[:]) 278 | cc[x].append('DELETE') 279 | if not subset: 280 | addtocc.setdefault(str(w), conjuncts) 281 | for w in cc.keys(): 282 | if 'DELETE' in cc[w]: 283 | del cc[w] 284 | for w,l in addtocc.iteritems(): 285 | cc.setdefault(w,l) 286 | 287 | # Sort all the lists so we can get the new grs right 288 | for w,l in cc.iteritems(): 289 | l.sort() 290 | 291 | # Add the new grs 292 | for w in cc.keys(): 293 | first_conjunct = cc[w][0] 294 | for c in cc[w][1:]: 295 | self.grs.append(GR('(conj %s %s)' % (first_conjunct, c), self.words)) 296 | # Add the cc 297 | newgr = GR('(cc %s %s)' % (first_conjunct, w), self.words) 298 | self.grs.append(newgr) 299 | 300 | # Finally, normalize any other grs in which all members of a 301 | # coordination participate - keep only the gr with the first conjunct 302 | grs_left = {} 303 | grs_right = {} 304 | for gr in self.grs: 305 | keyleft = str(gr.label)+" "+str(gr.args[0]) 306 | grs_left.setdefault(keyleft, []).append(gr.args[1]) 307 | keyright = str(gr.label)+" "+str(gr.args[1]) 308 | grs_right.setdefault(keyright, []).append(gr.args[0]) 309 | for k,l in grs_left.iteritems(): 310 | for w,c in cc.iteritems(): 311 | if k != w and is_subset(c,l): 312 | c.sort() 313 | labelparts = k.split() 314 | for item in c[1:]: 315 | testgr = GR('(%s %s %s)' % (labelparts[0], labelparts[1], item), self.words) 316 | for gr in self.grs: 317 | if gr.label == testgr.label and gr.args[0] == testgr.args[0] and gr.args[1] == testgr.args[1]: 318 | gr.ignore = True 319 | 320 | for k,l in grs_right.iteritems(): 321 | equal = False 322 | for w,c in cc.iteritems(): 323 | if k!= w and is_subset(c,l): 324 | c.sort 325 | labelparts = k.split() 326 | for item in c[1:]: 327 | testgr = GR('(%s %s %s)' % (labelparts[0], item, labelparts[1]), self.words) 328 | for gr in self.grs: 329 | if gr.label == testgr.label and gr.args[0] == testgr.args[0] and gr.args[1] == testgr.args[1]: 330 | gr.ignore = True 331 | 332 | 333 | def conj_to_appos(self): 334 | """ converts the pair of dependencies 335 | conj x y 336 | cc x , 337 | to the dependency: 338 | appos x y 339 | 340 | For example "x, a y, was found in ..." 341 | Takes advantage of the fact that the parser considers this 342 | a conjunction, and the merge_conj_args routine leaves it in this 343 | state because it never finds a conjunction word. Essentially 344 | this type of appos is like a conjunction without a conj word, 345 | so the comma is left hanging as a cc. 346 | 347 | Also takes advantage of the fact that the previous routine 348 | has already normalized GRs in which these things appear, so 349 | only the first element appears, which 350 | would have needed to be done for appositives anyway 351 | 352 | NB must be run AFTER merge_conj_args 353 | """ 354 | for gr in self.grs: 355 | if gr.label == 'cc' and gr.args[1].token == ',': 356 | gr.ignore = True 357 | for gr2 in self.grs: 358 | if gr2.label == 'conj' and gr2.args[0] == gr.args[0]: 359 | # could go wrong if there is something like "x, a y, and z" where there'd also be a (conj x z) relation that would get changed to appos. would require some fairly complicated checking but could be done. 360 | gr2.label = 'appos' 361 | 362 | def remove_comma_conj(self): 363 | for w in self.sentence: 364 | if w.token == ',' and w.grs: 365 | conj = [gr for gr in w.grs if gr.label == 'conj'] 366 | for gr in conj: 367 | gr.ignore = True 368 | 369 | if len(conj) == 2: 370 | head, mod = sorted([conj[0].args[1], conj[1].args[1]]) 371 | 372 | mod.sub = head 373 | head_grs = set([str(gr) for gr in head.grs]) 374 | for gr in mod.grs: 375 | if str(gr) in head_grs: 376 | gr.ignore = True 377 | mod.sub = None 378 | 379 | # self.grs.append(GR('(ncmod _ %s %s)' % (head, mod), self.words)) 380 | 381 | def num_and_number(self): 382 | """ converts nmod, amod, advmod, dep to dependency type num when the 383 | modifier is a number, i.e. has pos=CD 384 | or to dependency type number when both args are numbers 385 | """ 386 | deptypes = ['nmod', 'advmod', 'amod', 'dep'] 387 | for gr in self.grs: 388 | if gr.label in deptypes and gr.args[1].pos == 'CD': 389 | if gr.args[0].pos == 'CD': 390 | gr.label = 'number' 391 | gr.args[0], gr.args[1] = gr.args[1], gr.args[0] 392 | else: 393 | gr.label = 'num' 394 | 395 | def in_vivo_vitro(self): 396 | """ converts 397 | pobj in vivo/vitro 398 | to 399 | dep in vivo/vitro 400 | to match BioInfer. 401 | """ 402 | words = ['vivo', 'vitro'] 403 | for gr in self.grs: 404 | if gr.args[1].token in words and gr.args[0].token == 'in' and gr.label == 'pobj': 405 | gr.label = 'dep' 406 | 407 | def only_dep(self): 408 | """ converts 409 | advmod x only where only cat = ((S\NP)\(S\NP))/((S\NP)\(S\NP)) 410 | to 411 | dep x only 412 | to match BioInfer 413 | """ 414 | for gr in self.grs: 415 | if gr.label == 'advmod' and gr.args[1].token.lower() == 'only' and gr.args[1].cat == '((S\NP)\(S\NP))/((S\NP)\(S\NP))': 416 | gr.label = 'dep' 417 | 418 | def prep_to_appos(self): 419 | """ converts the pair of dependencies: 420 | pobj -LRB- x 421 | prep y -LRB- 422 | to the dependency: 423 | appox y x 424 | because parentheses are done as appos in BioInfer/SD. 425 | Should be possible to handle in markedup file, but 426 | not working at the moment. 427 | Also handles the case: 428 | nsubj y -LRB- 429 | rcmod x z 430 | to the dep: 431 | appos x y 432 | which comes from the category (NP\NP)/(S[dcl]\NP) just in case 433 | this is actually a parenthesis and not a relative pronoun. 434 | Also should be possible to handle in markedup file. 435 | This isn't perfect b/c if there are two rcmods it's indeterminate 436 | which one it will get - need to do some checking on the index. 437 | """ 438 | for gr in self.grs: 439 | if gr.label == 'prep' and ('-LRB-' in str(gr.args[1]) or '(' in str(gr.args[1])) and not gr.ignore: 440 | for gr2 in self.grs: 441 | if gr2.label == 'pobj' and gr.args[1] == gr2.args[0] and not gr2.ignore: 442 | gr.ignore = True 443 | self.grs.append(GR('(appos %s %s)' % (gr.args[0], gr2.args[1]), self.words)) 444 | gr2.ignore = True 445 | # note there may be more than one of these if another subroutine has modified the list - some may be ignore - just let it loop through all of them and not worry about it 446 | 447 | 448 | 449 | # The rcmod code isn't fully working 450 | for gr in self.grs: 451 | if gr.label == 'nsubj' and '-LRB-' in str(gr.args[1]): 452 | for gr2 in self.grs: 453 | if gr2.label == 'rcmod': 454 | gr.ignore = True 455 | gr2.ignore = True 456 | newgr = GR('(appos %s %s)' % (gr2.args[1],gr.args[0]), self.words) 457 | self.grs.append(newgr) 458 | break 459 | 460 | def switch_n_and_num(self): 461 | """ in our corpus if there is something like "residues 260-281" 462 | we may have residues|N/N 260-281|N, i.e. we may have the last 463 | item in the NP as the head, even if it is a number. 464 | however BioInfer has the number modifying the head 465 | as a num dependency. This switches those. 466 | If it finds: 467 | nmod y|CD x 468 | then it changes that to: 469 | num x y 470 | and anything of the form: 471 | reltype y z => reltype x z 472 | reltype z y => reltype z x 473 | 474 | NB needs to be run before prep_to_appos and nn_and_amond 475 | """ 476 | nums = [w for w in self.sentence if w.pos == 'CD'] 477 | for w in nums: 478 | prev = self.sentence[w.index - 1] 479 | if prev.cat == 'N/N' and w.cat == 'N': 480 | for gr in w.grs: 481 | if gr.label == 'nmod' and gr.args[0] == w and gr.args[1] == prev: 482 | # nmod y x => num x y 483 | gr.label = 'num' 484 | gr.args[0], gr.args[1] = gr.args[1], gr.args[0] 485 | elif gr.args[0] == w and gr.args[1] != prev: 486 | # reltype y z => reltype x z 487 | gr.ignore = True 488 | self.grs.append(GR('(%s %s %s)' % (gr.label, prev, gr.args[1]), self.words)) 489 | elif gr.args[1] == w and gr.args[0] != prev: 490 | # reltype z y => reltype z x 491 | gr.ignore = True 492 | self.grs.append(GR('(%s %s %s)' % (gr.label, gr.args[0], prev), self.words)) 493 | 494 | def fix_aux_conj(self): 495 | """ markedup_sd is set up so that aux+adj, aux+noun, aux+bare, aux+pass 496 | i.e. (s[dcl]\NP)/(S[adj]\NP) and (S[dcl]\NP)/NP 497 | and (S[dcl]\NP)/(S[b]\NP) and (S[dcl]\NP)/(S[pss]\NP) where the S[dcl] 498 | is an auxiliary, will give the adj/noun/bare as a head. 499 | However, in case those items are conjoined this doesn't work - 500 | the conjunction will still come out with the aux. 501 | This function fixes that. 502 | It also switches the head for some other relation types where we know it should be switched. (Not sure whether this should be everything but cop and aux, but playing it safe, just doing the ones we saw in dev data.) 503 | It relies on the raw parser output for conjunctions, so 504 | it must be run BEFORE merge_conj_args(). 505 | """ 506 | aux = ['ai', 'am', 'are', 'be', 'been', 'being', 'is', 'was', 'were', "'s", "'m", "'re", 'has', 'have', 'had', "'ve", 'do', 'did', 'does', "'d", "'ll", 'ca', 'can', 'could', 'may', 'might', 'must', 'ought', 'shall', 'should', 'will', 'wo', 'would', 'get', 'got', 'gotten', 'getting'] 507 | auxcats = ['(S[dcl]\NP)/(S[adj]\NP)', '(S[dcl]\NP)/NP', '(S[dcl]\NP)/(S[b]\NP)', '(S[dcl]\NP)/(S[pss]\NP)'] 508 | reltypes = ['ccomp', 'complm', 'rel', 'rcmod', 'advmod', 'partmod'] 509 | for gr in self.grs: 510 | # This is before we do merge_conj_args(), so the conjunction will be 511 | # args[0] and the aux will be args[1] 512 | if gr.label == 'conj' and gr.args[1].token in aux and gr.args[1].cat in auxcats: 513 | # Find the word that we need to switch in for the aux - 514 | # it will be the one which has this aux as its copula 515 | word = None 516 | for gr2 in self.grs: 517 | if (gr2.label == 'cop' or gr2.label == 'aux') and gr2.args[1] == gr.args[1]: 518 | word = gr2.args[0] 519 | if word: 520 | # change the other relations besides conj first 521 | for gr3 in self.grs: 522 | # could be on either side of the relation 523 | if gr3.label in reltypes and gr3.args[0] == gr.args[1]: 524 | gr3.args[0] = word 525 | if gr3.label in reltypes and gr3.args[1] == gr.args[1]: 526 | gr3.args[1] = word 527 | # now change the original conj relation 528 | gr.args[1] = word 529 | 530 | def partmod_to_xcomp(self): 531 | """ in cases like: 532 | suggested to act, data is presented to suggest, suggested to be involved, 533 | suggested to be a common region, proposed to act, predicted to encode, ... 534 | which always has the category sequence S[pss]\NP (S[to]\NP)/(S[b]\NP) 535 | the parser uses a unary rule and produces 536 | partmod x y 537 | This needs to be changed to 538 | xcomp x z 539 | where z is the complement of "to", which can be found because there 540 | is a relation (aux z to). 541 | """ 542 | for gr in self.grs: 543 | if gr.label == 'partmod' and gr.args[0].cat == 'S[pss]\NP' and gr.args[1].cat == '(S[to]\NP)/(S[b]\NP)': 544 | # Find the replacement word. 545 | # It will be the one which has this "to" as its aux. 546 | word = None 547 | for gr2 in self.grs: 548 | if gr2.label == 'aux' and gr2.args[1] == gr.args[1]: 549 | word = gr2.args[0] 550 | if word: 551 | gr.label = 'xcomp' 552 | gr.args[1] = word 553 | 554 | def ecm_to_nsubj(self): 555 | """ for ECM (and control) constructions like this: 556 | these|DT|NP[nb]/N reactions|NNS|N may|MD|(S[dcl]\NP)/(S[b]\NP) 557 | allow|VB|((S[b]\NP)/(S[to]\NP))/NP profilin|NN|N 558 | to|TO|(S[to]\NP)/(S[b]\NP) have|VB|(S[b]\NP)/NP 559 | These come out with (xsubj have profilin) and it should 560 | be nsubj. 561 | We deliberately left (dobj allow profilin) in the markedup 562 | file, even though Stanford doesn't want that, to have a 563 | hook to fix this construction with. 564 | So, this subroutine changes the pair: 565 | (dobj x y) where cat x ends in /NP 566 | (xsubj z y) 567 | to 568 | (nsubj z y) 569 | """ 570 | verbs = [w for w in self.sentence if w.cat.endswith('/NP')] 571 | for v in verbs: 572 | grs = [gr for gr in v.grs if gr.args[0] == v and gr.label == 'dobj'] 573 | for gr in grs: 574 | for gr2 in self.grs: 575 | if gr2.label == 'xsubj' and gr2.args[1] == gr.args[1]: 576 | gr.ignore = True 577 | gr2.label = 'nsubj' 578 | 579 | def preconj(self): 580 | """ converts 581 | advmod x both / advmod x either 582 | to 583 | preconj x both / preconj x either 584 | """ 585 | preconjwords = ['both', 'either'] 586 | for gr in self.grs: 587 | if gr.label == 'advmod' and gr.args[1].token.lower() in preconjwords: 588 | gr.label = 'preconj' 589 | 590 | def less_than(self): 591 | """ handles the less than construction in BioInfer. 592 | specifically if the sentence contains 593 | x < n (where n has pos = CD) 594 | then it changes the pair of dependencies: 595 | num < n 596 | nmod < x 597 | to the pair: 598 | num x n 599 | dep x < 600 | to match BioInfer. Then, since BioInfer considers the head 601 | of this construction to be x, and our script has made it <, 602 | it changes < in any other relation to x. 603 | 604 | NB must be run after switch_n_and_num() 605 | """ 606 | 607 | lts = [w for w in self.sentence if w.token == '<' and w.cat == 'N/N'] 608 | for w in lts: 609 | prev = self.sentence[w.index - 1] 610 | next = self.sentence[w.index + 1] 611 | if prev.cat == 'N/N' and next.pos == 'CD' and next.cat == 'N': 612 | for gr in w.grs: 613 | if gr.label == 'num' and gr.args[0] == w and gr.args[1] == next: 614 | gr.args[0] = prev 615 | elif gr.label == 'nmod' and gr.args[0] == w and gr.args[1] == prev: 616 | gr.label = 'dep' 617 | gr.args[0], gr.args[1] = gr.args[1], gr.args[0] 618 | elif gr.args[0] == w and gr.args[1] != prev: 619 | # reltype < z => reltype x z 620 | gr.ignore = True 621 | self.grs.append(GR('(%s %s %s)' % (gr.label, prev, gr.args[1]), self.words)) 622 | elif gr.args[1] == w and gr.args[0] != prev: 623 | # reltype z < => reltype z x 624 | gr.ignore = True 625 | self.grs.append(GR('(%s %s %s)' % (gr.label, gr.args[0], prev), self.words)) 626 | 627 | 628 | 629 | def fix_ncmod_num(self): 630 | for w in self.cats.get('N[num]', []): 631 | for gr in w.grs: 632 | if gr.label != 'ncmod': 633 | continue 634 | if gr.args[1] == w: 635 | gr.args[0] = 'num' 636 | gr.args[1], gr.args[2] = gr.args[2], gr.args[1] 637 | elif gr.args[2] == w: 638 | gr.args[0] = 'num' 639 | gr.args[2] = self.sentence[gr.args[1].index + 1] 640 | for gr in self.grs: 641 | if gr.label == 'ncmod' and \ 642 | (gr.args[1].is_num() or gr.args[2].is_num()): 643 | gr.args[0] = 'num' 644 | 645 | PART = set("several some most all any more less many".split()) 646 | def ncmod_to_iobj(self): 647 | for gr in self.grs: 648 | if gr.label == 'ncmod' and gr.args[2].token == 'of': 649 | if gr.args[1].token.lower() not in self.PART: 650 | gr.args.pop(0) 651 | gr.label = 'iobj' 652 | 653 | def sfs_to_conj(self): 654 | for gr in self.grs: 655 | if gr.label == 'ncmod' and gr.args[2].pos == 'CC' and gr.args[2].cat == 'S/S': 656 | self.grs.append(GR('(conj %s %s)' % (gr.args[2], gr.args[1]), self.words)) 657 | gr.ignore = True 658 | 659 | def pct_ncmod_to_dobj(self): 660 | for gr in self.grs: 661 | if gr.label == 'ncmod' and gr.args[0] == '_' and gr.args[2].token == '%': 662 | gr.args.pop(0) 663 | gr.label = 'dobj' 664 | 665 | def xmod_to_ta(self): 666 | for gr in self.grs: 667 | if gr.label == 'xmod': 668 | start, end = sorted([gr.args[1].index, gr.args[2].index]) 669 | for i in xrange(start + 1, end): 670 | if self.sentence[i].token == ',': 671 | gr.label = 'ta' 672 | for i in xrange(end + 1, len(self.sentence)): 673 | if self.sentence[i].token == ',': 674 | gr.args[0] = 'bal' 675 | break 676 | else: 677 | gr.args[0] = 'end' 678 | break 679 | 680 | def xmod_add_to(self): 681 | for gr in self.grs: 682 | if gr.label == 'xmod': 683 | start, end = sorted([gr.args[1].index, gr.args[2].index]) 684 | for i in xrange(start + 1, end): 685 | if self.sentence[i].token == 'to': 686 | gr.args[0] = 'to' 687 | break 688 | 689 | SAY = set("say said says".split()) 690 | def cmod_to_ta(self): 691 | for gr in self.grs: 692 | if gr.label == 'cmod': 693 | if gr.args[2].token.lower() in self.SAY: 694 | gr.label = 'ta' 695 | gr.args[0] = 'quote' 696 | gr.args[1], gr.args[2] = gr.args[2], gr.args[1] 697 | elif gr.args[1].token.lower() in self.SAY: 698 | gr.label = 'ta' 699 | gr.args[0] = 'quote' 700 | 701 | def ncmod_to_prt(self): 702 | for gr in self.grs: 703 | if gr.label == 'ncmod': 704 | if gr.args[1].pos.startswith('V') and gr.args[2].pos == 'RP': 705 | gr.args[0] = 'prt' 706 | 707 | def filter_punct(self): 708 | for gr in self.grs: 709 | for arg in gr.args: 710 | if isinstance(arg, Word) and arg.token in PUNCT: 711 | gr.ignore = True 712 | break 713 | 714 | def fix_ampersands(self): 715 | for w in self.sentence: 716 | if w.token == '&' and w.cat == 'N/N': 717 | prev = self.sentence[w.index - 1] 718 | next = self.sentence[w.index + 1] 719 | for gr in w.grs: 720 | gr.ignore = True 721 | for gr in next.grs: 722 | if prev in gr.args: 723 | gr.ignore = True 724 | else: 725 | gr.replace(next, w) 726 | self.grs.append(GR('(conj %s %s)' % (w, prev), self.words)) 727 | self.grs.append(GR('(conj %s %s)' % (w, next), self.words)) 728 | 729 | def remove_relpro_cmod(self): 730 | for gr in self.grs: 731 | if gr.label == 'cmod' and gr.args[0] != '_' and gr.args[0].token != 'that': 732 | gr.args[0] = '_' 733 | 734 | def fix_that_relpro(self): 735 | for gr in self.grs: 736 | if gr.label == "cmod" and gr.args[0] != '_' and gr.args[0].token == 'that': 737 | that = gr.args[0] 738 | for gr2 in that.grs: 739 | if gr2.label == "ncsubj" and gr2.args[1] == that: 740 | gr2.args[1] = gr.args[1] 741 | 742 | def add_rel(self): 743 | """ in case of these two deps: 744 | rcmod x y 745 | nsubj(pass) y that/which 746 | adds an ADDITIONAL dep: 747 | rel y that/which 748 | In other words, adds a dependency to mark the relative 749 | pronoun as being rel. Note that in BioInfer there are 750 | BOTH the nsubj and the rel dependencies. Haven't checked 751 | whether this is the case for Stanford parser in general. 752 | also adds a second ADDITIONAL dep: 753 | ref x that/which 754 | this can't be done in the markedup file because would be 755 | one-to-many for (NP\NP)/(S[dcl]\NP) 756 | """ 757 | for gr in self.grs: 758 | if gr.label == 'rcmod': 759 | verb = gr.args[1] 760 | for gr2 in verb.grs: 761 | if (gr2.label == 'nsubj' or gr2.label == 'nsubjpass') and gr2.args[0] == verb and (gr2.args[1].token == 'that' or gr2.args[1].token == 'which'): 762 | self.grs.append(GR('(rel %s %s)' % (gr2.args[0], gr2.args[1]), self.words)) 763 | self.grs.append(GR('(ref %s %s)' % (gr.args[0], gr2.args[1]), self.words)) 764 | 765 | def read(filename, OUTPUT): 766 | grs = [] 767 | for line in open(filename): 768 | line = line.strip() 769 | if not line or line.startswith('#'): 770 | if OUTPUT == '--ccgbank': 771 | print line 772 | continue 773 | # line = punct_convert(line) 774 | if line.startswith('('): 775 | grs.append(line) 776 | elif line.startswith(' '): 777 | yield Sentence(grs, line[4:]) 778 | grs = [] 779 | 780 | OUTPUT = sys.argv[1] 781 | if sys.argv[2].startswith('--'): 782 | if sys.argv[2] != '--no-postprocess': 783 | print >> sys.stderr, "unrecognised command line argument %s" % sys.argv[2] 784 | sys.exit(1) 785 | 786 | POSTPROCESS = False 787 | FILENAME = sys.argv[3] 788 | else: 789 | POSTPROCESS = True 790 | FILENAME = sys.argv[2] 791 | 792 | if OUTPUT == "--ccgbank": 793 | print "# generated by grs2sd" 794 | elif OUTPUT == '--rasp-parse': 795 | print "%LB (" 796 | print "%RB )" 797 | 798 | for i, s in enumerate(read(FILENAME, OUTPUT)): 799 | if POSTPROCESS: 800 | s.postprocess() 801 | 802 | if OUTPUT == '--ccgbank': 803 | for gr in s.grs: 804 | if not gr.ignore: 805 | print gr 806 | 807 | print '', 808 | for w in s.sentence: 809 | print '%s|%s|%s' % (w.token, w.pos, w.cat), 810 | print 811 | elif OUTPUT == '--rasp-text': 812 | print 813 | print i + 1 814 | for w in s.sentence: 815 | print w.token, 816 | print 817 | elif OUTPUT == '--rasp-parse': 818 | print 819 | print i + 1 820 | print 821 | for gr in s.grs: 822 | if not gr.ignore: 823 | print gr.nopos() 824 | elif OUTPUT == '--text': 825 | for w in s.sentence: 826 | print w.token, 827 | print 828 | elif OUTPUT == '--pos': 829 | for w in s.sentence: 830 | print '%s|%s' % (w.token, w.pos), 831 | print 832 | -------------------------------------------------------------------------------- /downloads/markedup_sd-1.00: -------------------------------------------------------------------------------- 1 | # C&C NLP tools 2 | # Copyright (c) Universities of Edinburgh, Oxford and Sydney 3 | # Copyright (c) James R. Curran 4 | # 5 | # This software is covered by a non-commercial use licence. 6 | # See LICENCE.txt for the full text of the licence. 7 | # 8 | # If LICENCE.txt is not included in this distribution 9 | # please email candc@it.usyd.edu.au to obtain a copy. 10 | # This is the 'markedup' file which maps from bare categories to the 11 | # categories marked up with head and dependency information. This 12 | # file now also contains the mapping to grammatical relations based 13 | # on the Briscoe and Carroll scheme. 14 | # 15 | # Editing existing marked up categories in this file without regenerating 16 | # the training data and reestimating the model will lead to reduced 17 | # performance. However, it is possible to change the mapping to grammatical 18 | # relations without affecting the existing model. 19 | # 20 | # The category information is in the following format: 21 | # 22 | # <#_of_slots> 23 | # [! ] 24 | # 25 | # 26 | # ... 27 | # 28 | # The exclamation marked entry is an alternative markedup category 29 | # with a different head passing scheme to match Briscoe and Carroll. 30 | # 31 | # The slots in the CCG markup aren't always numbered from left to right (for 32 | # historical reasons). This is an inconsistency we will fix in a later version. 33 | # 34 | # The gr annotation is based on the Briscoe and Carroll (B+C) GR scheme. 35 | # See the following technical report for details: 36 | # 37 | # Ted Briscoe 38 | # An Introduction to Tag Sequence Grammars and the RASP System Parser 39 | # Technical Report: UCAM-CL-TR-662 40 | # University of Cambridge Computing Laboratory 41 | # 42 | # Some things to fix: 43 | # 44 | # the subject in (S[adj]\NP)/NP gets ignored, which is correct except when the dependency has 45 | # come about through a unary rule. So we're currently not producing this unary rule dependency. 46 | # (There may be other cases like this.) 47 | # 48 | # The GR markup for the question categories at the end of the file needs looking at again 49 | # 50 | # The CCG dependencies haven't been changed, even where there are obvious improvements, 51 | # because this version of the markedup file was used to get the CL-paper results and 52 | # we'd like this to serve as part of the record of those results. 53 | 54 | # lexical constraint groups 55 | =be ai am are be been being is was were 's 'm 're 56 | =aux ai am are be been being is was were 's 'm 're 57 | =aux has have had 've 58 | =aux do did does 'd 'll ca can could may might must ought 59 | =aux shall should will wo would 60 | =aux get got gotten getting 61 | =det another other some such 62 | =fulldet the this that a an 63 | =part all both 64 | =neg not n't 65 | =bracket -LRB- -RRB- ( ) -LCB- -RCB- { } -LSB- -RSB- [ ] 66 | # now list the markedup categories 67 | , 68 | 0 ,{_} 69 | 70 | . 71 | 0 .{_} 72 | 73 | : 74 | 0 :{_} 75 | 76 | ; 77 | 0 ;{_} 78 | 79 | LRB 80 | 0 LRB{_} 81 | 82 | N 83 | 0 N{_} 84 | 85 | NP 86 | 0 NP{_} 87 | 88 | NP[expl] 89 | 0 NP[expl]{_} 90 | 91 | PP 92 | 0 PP{_} 93 | 94 | RRB 95 | 0 RRB{_} 96 | 97 | S 98 | 0 S{_} 99 | 100 | S[dcl] 101 | 0 S[dcl]{_} 102 | 103 | S[frg] 104 | 0 S[frg]{_} 105 | 106 | S[intj] 107 | 0 S[intj]{_} 108 | 109 | S[wq] 110 | 0 S[wq]{_} 111 | 112 | conj 113 | 0 conj{_} 114 | 115 | N[num] 116 | 0 N[num]{_} 117 | 118 | NP[thr] 119 | 0 NP[thr]{_} 120 | 121 | N/N 122 | 1 (N{Y}/N{Y}<1>){_} 123 | 1 nmod %f %l # BioInfer merges nn and amod into nmod 124 | 125 | NP[nb]/N 126 | 1 (NP[nb]{Y}/N{Y}<1>){_} 127 | 1 det %f %l # SD doesn't distinguish "both/all" as ncmod part. 128 | 129 | (NP\NP)/NP 130 | 2 ((NP{Y}\NP{Y}<1>){_}/NP{Z}<2>){_} 131 | 1 prep %f %l 132 | 2 pobj %l %f 133 | 134 | ((S\NP)\(S\NP))/NP 135 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 136 | 1 prep %f %l 137 | 2 pobj %l %f 138 | 139 | PP/NP 140 | 1 (PP{_}/NP{Y}<1>){_} 141 | 1 pobj %l %f 142 | 143 | (S[dcl]\NP)/NP 144 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 145 | 1 nsubj %2 %f =be 146 | 1 nsubj %l %f 147 | 2 cop %f %l =be 148 | 2 dobj %l %f 149 | 150 | (S\NP)\(S\NP) 151 | 1 ((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_} 152 | 1 neg %f %l =neg 153 | 1 advmod %f %l 154 | 155 | (S[b]\NP)/NP 156 | 2 ((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 157 | 1 ignore 158 | 2 cop %f %l =be 159 | 2 dobj %l %f 160 | 161 | (S[to]\NP)/(S[b]\NP) 162 | 2 ((S[to]{_}\NP{Z}<1>){_}/(S[b]{Y}<2>\NP{Z*}){Y}){_} 163 | 1 xsubj %2 %f 164 | 2 aux %f %l 165 | 166 | (S[dcl]\NP)/(S[b]\NP) 167 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 168 | ! ((S[dcl]{Z}\NP{Y}<1>){Z}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 169 | 1 nsubj %2 %f =aux 170 | 1 nsubj %l %f 171 | 2 aux %f %l =aux 172 | 2 xcomp %l %f 173 | 174 | (NP[nb]/N)\NP 175 | 2 ((NP[nb]{Y}/N{Y}<1>){_}\NP{Z}<2>){_} 176 | 1 poss %f %2 177 | 2 possessive %f %l 178 | 179 | S[adj]\NP 180 | 1 (S[adj]{_}\NP{Y}<1>){_} 181 | 1 nsubj %l %f 182 | 183 | S[pss]\NP 184 | 1 (S[pss]{_}\NP{Y}<1>){_} 185 | 1 nsubjpass %l %f 186 | 187 | (N/N)/(N/N) 188 | 1 ((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}<1>){_} 189 | 1 advmod %f %l 190 | 191 | (S[ng]\NP)/NP 192 | 2 ((S[ng]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 193 | 1 nsubj %l %f 194 | 2 dobj %l %f 195 | 196 | (S\NP)/(S\NP) 197 | 1 ((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_} 198 | 1 advmod %f %l 199 | 200 | (S[dcl]\NP)/S[dcl] 201 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 202 | 1 nsubj %l %f 203 | 2 ccomp %l %f 204 | 205 | S[dcl]\NP 206 | 1 (S[dcl]{_}\NP{Y}<1>){_} 207 | 1 nsubj %l %f 208 | 209 | (S[dcl]\NP)/(S[pt]\NP) 210 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 211 | ! ((S[dcl]{Z}\NP{Y}<1>){Z}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 212 | 1 ignore =aux 213 | 1 nsubj %l %f 214 | 2 aux %f %l =aux 215 | 2 xcomp %l %f 216 | 217 | S/S 218 | 1 (S[X]{Y}/S[X]{Y}<1>){_} 219 | 1 advmod %f %l 220 | 221 | (NP\NP)/(S[dcl]\NP) 222 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[dcl]{Z}<2>\NP{Y*}){Z}){_} 223 | ! ((NP{Y}\NP{Y}<1>){_}/(S[dcl]{Z}<2>\NP{_}){Z}){_} 224 | 1 rcmod %f %2 225 | 2 ignore 226 | 227 | (S/S)/NP 228 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/NP{Z}<2>){_} 229 | 1 prep %f %l 230 | 2 pobj %l %f 231 | 232 | (S[dcl]\NP)/(S[adj]\NP) 233 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 234 | ! ((S[dcl]{Z}\NP{Y}<1>){Z}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 235 | 1 ignore 236 | 2 cop %f %l # Stanford parser considers the verb to be a copula even when it's not an aux 237 | 238 | (S[dcl]\NP)/(S[pss]\NP) 239 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 240 | ! ((S[dcl]{Z}\NP{Y}<1>){Z}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 241 | 1 ignore =aux 242 | 1 nsubj %l %f 243 | 2 auxpass %f %l =aux 244 | 2 cop %f %l 245 | 246 | S[em]/S[dcl] 247 | 1 (S[em]{_}/S[dcl]{Y}<1>){_} 248 | 1 complm %f %l 249 | 250 | ((S\NP)\(S\NP))/((S\NP)\(S\NP)) 251 | 1 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_} 252 | 1 advmod %f %l # Is advmod most of the time in BioInfer dev set 253 | 254 | (S[dcl]\NP)/(S[ng]\NP) 255 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 256 | ! ((S[dcl]{Z}\NP{Y}<1>){Z}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 257 | 1 ignore =aux 258 | 1 nsubj %l %f 259 | 2 aux %f %l =aux # actually need different condition if arg is (S[ng]\NP)/(S[adj]\NP) and an aux - see "She is being prudent" 260 | 2 partmod %l %f 261 | 262 | ((S\NP)\(S\NP))/S[dcl] 263 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S[dcl]{W}<2>){_} 264 | 1 advcl %f %2 # not sure this is correct all the time 265 | 2 mark %f %l 266 | 267 | (S[dcl]\NP)/PP 268 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 269 | 1 nsubj %l %f 270 | 2 prep %l %f =PP/NP # BioInfer doesn't distinguish complement from adjunct PPs 271 | 2 advcl %l %c # e.g. had received ten letters since the changes were proposed relation between received and propsed 272 | # Right now no multiple tests, only one test and one elsewhere case 273 | # 2 xcomp _ %l %f =PP/(S[ng]\NP) 274 | # 2 xcomp _ %l %f =PP/(S[adj]\NP) 275 | # 2 ccomp _ %l %f =PP/S[dcl] 276 | 277 | (S[dcl]\NP)/(S[to]\NP) 278 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 279 | 1 nsubj %l %f 280 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 281 | 282 | S[b]\NP 283 | 1 (S[b]{_}\NP{Y}<1>){_} 284 | 1 ignore 285 | 286 | NP\NP 287 | 1 (NP{Y}\NP{Y}<1>){_} 288 | 1 mod %f %l # unfortunately this could be anything, tmod, advmod, ... 289 | 290 | (S[pt]\NP)/NP 291 | 2 ((S[pt]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 292 | 1 nsubj %l %f 293 | 2 cop %f %l =be 294 | 2 dobj %l %f 295 | 296 | (S[pss]\NP)/PP 297 | 2 ((S[pss]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 298 | 1 nsubjpass %l %f 299 | 2 advmod %l %f =PP/(S[adj]\NP) # "characterized as immoral" - Stanford parser has odd analysis for this - also you'd want dep(characterized,immoral) which I can't represent here. 300 | 2 prep %l %f 301 | # 2 prep %l %f =PP/NP # BioInfer doesn't distinguish complement from adjunct PPs 302 | # 2 prep %l %f =PP/(S[ng]\NP) # "was engaged in pulling out the" Stanford parser treats "in" as a prep, incorrectly I think since it probably should be ccomp or xcomp 303 | 304 | (S[dcl]\NP)/S[em] 305 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 306 | 1 nsubj %l %f 307 | 2 ccomp %l %c =S[em]/S[dcl] # correct, but not sure why these two arg types are singled out 308 | 2 ccomp %l %c =S[em]/S[b] # correct, but ditto 309 | 310 | (S[adj]\NP)/(S[adj]\NP) 311 | 1 ((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}<1>\NP{Z}){Y}){_} 312 | 1 advmod %f %l 313 | 314 | NP/NP 315 | 1 (NP{Y}/NP{Y}<1>){_} 316 | 1 det %f %l =det 317 | 1 advmod %f %l 318 | 319 | (S/S)/S[dcl] 320 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/S[dcl]{Z}<2>){_} 321 | 1 advcl %f %2 322 | 2 mark %f %l 323 | 324 | (S[dcl]\S[dcl])\NP 325 | 2 ((S[dcl]{_}\S[dcl]{Y}<1>){_}\NP{Z}<2>){_} 326 | 1 ccomp %l %f 327 | 2 nsubj %l %f 328 | # not sure about the order of %l %f on 1: doesn't seem to be consistent in B+C 329 | 330 | (S[b]\NP)/(S[pss]\NP) 331 | 2 ((S[b]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 332 | ! ((S[b]{Z}\NP{Y}<1>){Z}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 333 | 1 ignore 334 | 2 auxpass %f %l =aux 335 | 2 xcomp %l %f # hard to find an example 336 | 337 | (NP\NP)/(NP\NP) 338 | 1 ((NP{Y}\NP{Y}){Z}/(NP{Y}\NP{Y}){Z}<1>){_} 339 | 1 dep %f %l # "such as" Stanford parser has simply dep, although "in hearly summer next year" it has amod(year, next) so difficult. 340 | 341 | (S[b]\NP)/PP 342 | 2 ((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 343 | 1 ignore 344 | 2 prep %l %f =PP/NP # BioInfer doesn't distinguish complement from adjunct PPs 345 | 2 advcl %l %c 346 | # Right now no multiple tests, only one test and one elsewhere case 347 | # 2 xcomp _ %l %f =PP/(S[ng]\NP) 348 | # 2 xcomp _ %l %f =PP/(S[adj]\NP) 349 | # 2 ccomp _ %l %f =PP/S[dcl] 350 | 351 | (NP\NP)/N 352 | 2 ((NP{Y}\NP{Y}<1>){_}/N{Z}<2>){_} 353 | 1 prep %f %l 354 | 2 pobj %l %f 355 | 356 | S[ng]\NP 357 | 1 (S[ng]{_}\NP{Y}<1>){_} 358 | 1 nsubj %l %f 359 | 360 | ((S[dcl]\NP)/PP)/NP 361 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 362 | 1 nsubj %l %f 363 | 2 prep %l %f =PP/NP # BioInfer doesn't distinguish complement from adjunct PPs 364 | 2 xcomp %l %f =PP/(S[ng]\NP) # hard to find an example 365 | 2 xcomp %l %f =PP/(S[adj]\NP) # hard to find an example 366 | 3 dobj %l %f 367 | 368 | N\N 369 | 1 (N{Y}\N{Y}<1>){_} 370 | 1 dep %f %l # problem, could be anything from dep, to measure, to nsubj ... 371 | 372 | (S[adj]\NP)/PP 373 | 2 ((S[adj]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 374 | 1 nsubj %l %f 375 | 2 prep %l %f =PP/NP # BioInfer doesn't distinguish complement from adjunct PPs 376 | 2 xcomp %l %f =PP/(S[ng]\NP) # hard to find an example 377 | 378 | ((S[b]\NP)/PP)/NP 379 | 3 (((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 380 | 1 ignore 381 | 2 prep %l %f =PP/NP # BioInfer doesn't distinguish complement from adjunct PPs 382 | 2 xcomp %l %f =PP/(S[ng]\NP) # hard to find an example 383 | 2 xcomp %l %f =PP/(S[adj]\NP) # hard to find an example 384 | 3 dobj %l %f 385 | 386 | (S[dcl]\S[dcl])/NP 387 | 2 ((S[dcl]{_}\S[dcl]{Y}<1>){_}/NP{Z}<2>){_} 388 | 1 ccomp %l %f 389 | 2 nsubj %l %f 390 | 391 | ((S\NP)\(S\NP))/PP 392 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/PP{W}<2>){_} 393 | # 1 (pmod %f %l) POS==IN 394 | 1 mod %f %l # not clear what to do here; many different constructions in CCGbank 395 | 2 prep %l %f =PP/NP 396 | # do we want to support POS constraints (and pmod)? 397 | 398 | (S[b]\NP)/(S[adj]\NP) 399 | 2 ((S[b]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 400 | ! ((S[b]{Z}\NP{Y}<1>){Z}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 401 | 1 ignore 402 | 2 cop %f %l 403 | 404 | (S[ng]\NP)/PP 405 | 2 ((S[ng]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 406 | 1 nsubj %l %f 407 | 2 prep %l %f =PP/NP 408 | 2 advcl %l %f 409 | # Right now no multiple tests, only one test and one elsewhere case 410 | # 2 xcomp _ %l %f =PP/(S[ng]\NP) 411 | # 2 xcomp _ %l %f =PP/(S[adj]\NP) 412 | # 2 ccomp _ %l %f =PP/S[dcl] 413 | 414 | (S[pss]\NP)/(S[to]\NP) 415 | 2 ((S[pss]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 416 | 1 nsubjpass %l %f 417 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 418 | 419 | ((S\NP)\(S\NP))/(S[ng]\NP) 420 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[ng]{W}<2>\NP{Z*}){W}){_} 421 | 1 prep %f %l 422 | 2 pobj %l %f 423 | 424 | ((S\NP)\(S\NP))/N 425 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/N{W}<2>){_} 426 | 1 mod %f %2 # difficult, could be tmod or many others 427 | 2 det %f %l # e.g. "will be hit the hardest" or "was changed this week" 428 | 429 | (NP\NP)/S[dcl] 430 | 2 ((NP{Y}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 431 | 1 rcmod %f %2 432 | 2 dep %f %l # difficult: could be "that" (complm), "where" (advmod), ... 433 | 434 | (S[adj]\NP)/(S[to]\NP) 435 | 2 ((S[adj]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 436 | 1 nsubj %l %f 437 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 438 | 439 | PP/(S[ng]\NP) 440 | 1 (PP{_}/(S[ng]{Y}<1>\NP{Z}){Y}){_} 441 | 1 pobj %l %f 442 | 443 | (NP\NP)/(S[ng]\NP) 444 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 445 | ! ((NP{Y}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y}){Z}){_} 446 | 1 prep %f %l 447 | 2 pobj %l %f 448 | 449 | ((S[adj]\NP)\(S[adj]\NP))/NP 450 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 451 | 1 prep %f %l 452 | 2 pobj %l %f 453 | 454 | S[pt]\NP 455 | 1 (S[pt]{_}\NP{Y}<1>){_} 456 | 1 nsubj %l %f 457 | 458 | (S[ng]\NP)/(S[to]\NP) 459 | 2 ((S[ng]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 460 | 1 nsubj %l %f 461 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 462 | 463 | (S/S)/(S/S) 464 | 1 ((S[X]{Y}/S[X]{Y}){Z}/(S[X]{Y}/S[X]{Y}){Z}<1>){_} 465 | 1 advmod %f %l 466 | 467 | (S[b]\NP)/(S[to]\NP) 468 | 2 ((S[b]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 469 | 1 nsubj %l %f 470 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 471 | 472 | ((S[ng]\NP)/PP)/NP 473 | 3 (((S[ng]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 474 | 1 nsubj %l %f 475 | 2 prep %l %f =PP/NP 476 | 2 xcomp %l %f =PP/(S[ng]\NP) # hard to find an example 477 | 2 xcomp %l %f =PP/(S[adj]\NP) # hard to find an example 478 | 3 dobj %l %f 479 | 480 | ((S[dcl]\NP)/(S[to]\NP))/NP 481 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 482 | 1 nsubj %l %f 483 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 484 | 3 dobj %l %f # will get processed out by python script 485 | 486 | (S[pt]\NP)/(S[pss]\NP) 487 | 2 ((S[pt]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 488 | ! ((S[pt]{Z}\NP{Y}<1>){Z}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 489 | 1 ignore =aux 490 | 1 nsubj %l %f 491 | 2 auxpass %f %l =aux 492 | 2 xcomp %l %f 493 | 494 | S[qem]/S[dcl] 495 | 1 (S[qem]{_}/S[dcl]{Y}<1>){_} 496 | 1 advmod %f %l # I believe this is advmod when the lexical item is when/how/whether etc, but mark when the lexical item is if (and maybe others) - need to determine what the distribution is and set up equivalence classes 497 | 498 | ((S\NP)/(S\NP))/NP 499 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 500 | 1 prep %f %l 501 | 2 pobj %l %f 502 | 503 | N/S[em] 504 | 1 (N{_}/S[em]{Y}<1>){_} 505 | 1 ignore 506 | # 1 ccomp %f %l %c =S[em]/S[dcl] 507 | # I've commented out the original GR output line. I think Stanford handles this badly - the Stanford parser gives no relation between "disbelief" and "alive" (or "was") in the sentence "The community expressed disbelief that Joe was alive." 508 | # However there is a relation ccomp(expressed, alive) which I think needs to be handled on (S[x]\NP)/NP when the dobj is N/S[x], particularly N/S[em] but could be N/S[dcl]. This may not be possible until the bug with multiple conditions is fixed. 509 | 510 | (S[pss]\NP)/NP 511 | 2 ((S[pss]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 512 | 1 nsubjpass %l %f 513 | 2 dobj %l %f # "I was given a book" 514 | 515 | (S[adj]\NP)/NP 516 | 2 ((S[adj]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 517 | 1 nsubj %l %f 518 | 2 dobj %l %f 519 | 520 | (S[b]\NP)/(S[pt]\NP) 521 | 2 ((S[b]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 522 | ! ((S[b]{Z}\NP{Y}<1>){Z}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 523 | 1 ignore =aux 524 | 1 nsubj %l %f 525 | 2 aux %f %l =aux 526 | 2 xcomp %l %f # difficult to find an example 527 | 528 | (S[b]\NP)/(S[ng]\NP) 529 | 2 ((S[b]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 530 | ! ((S[b]{Z}\NP{Y}<1>){Z}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 531 | 1 ignore =aux 532 | 1 nsubj %l %f 533 | 2 aux %f %l =aux 534 | 2 partmod %l %f 535 | 536 | (NP\NP)/(S[dcl]/NP) 537 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[dcl]{Z}<2>/NP{Y*}){Z}){_} 538 | 1 ccomp %f %2 539 | 2 complm %l %f # complm if "that", dobj if "whom" etc. 540 | 541 | (S[pt]\NP)/PP 542 | 2 ((S[pt]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 543 | 1 nsubj %l %f 544 | 2 xcomp %l %c =PP/(S[adj]\NP) # not sure about this one 545 | 2 prep %l %f 546 | 547 | PP/PP 548 | 1 (PP{Y}/PP{Y}<1>){_} 549 | 1 dep %f %l # Stanford parser not very clear on what the options are here 550 | 551 | ((S[b]\NP)/(S[to]\NP))/NP 552 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 553 | 1 nsubj %l %f 554 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 555 | 3 dobj %l %f # will get processed out by python script 556 | 557 | (NP\NP)/PP 558 | 2 ((NP{Y}\NP{Y}<1>){_}/PP{Z}<2>){_} 559 | 1 dep %f %l # not very clear what the options are 560 | 2 dep %l %f =PP/NP # to match Stanford parser output 561 | 562 | (S[pt]\NP)/(S[ng]\NP) 563 | 2 ((S[pt]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 564 | ! ((S[pt]{Z}\NP{Y}<1>){Z}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 565 | 1 ignore =aux 566 | 1 nsubj %l %f 567 | 2 aux %f %l =aux 568 | 2 partmod %l %f 569 | 570 | (S/S)/PP 571 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/PP{Z}<2>){_} 572 | 1 advmod %f %l 573 | 2 dep %l %f =PP/NP # to match Stanford parser output 574 | 575 | (S[pt]\NP)/(S[adj]\NP) 576 | 2 ((S[pt]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 577 | 1 ignore 578 | 2 comp %f %l 579 | 580 | (S[ng]\NP)/(S[pss]\NP) 581 | 2 ((S[ng]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 582 | ! ((S[ng]{Z}\NP{Y}<1>){Z}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 583 | 1 ignore =aux 584 | 1 nsubj %l %f 585 | 2 auxpass %f %l =aux 586 | 2 xcomp %l %f 587 | 588 | conj/conj 589 | 1 (conj{Y}/conj{Y}<1>){_} 590 | 1 dep %f %l # this is to match Stanford parser, but the direction of the dependency is not consistent in the parser output 591 | 592 | (S[adj]\NP)\NP 593 | 2 ((S[adj]{_}\NP{Y}<1>){_}\NP{Z}<2>){_} 594 | 1 nsubj %l %f 595 | 2 mod %f %l # difficult, could be advmod "some distance apart", measure "three years old", ... 596 | 597 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/NP 598 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}<3>){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_}/NP{V}<2>){_} 599 | 1 ignore # "go higher than 2000 feet", "was down 4.5 from 70" - in SD the "than" or "from" is analyzed with a dependency (prep) with the verb (go/was), not the verbal modifier (higher/down). 600 | 2 pobj %l %f 601 | 3 prep %f %l # tried adding this argument 602 | 603 | ((S[dcl]\NP)/NP)/NP 604 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 605 | 1 nsubj %l %f 606 | 2 dobj %l %f 607 | 3 iobj %l %f 608 | 609 | (S\S)/S[dcl] 610 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/S[dcl]{Z}<2>){_} 611 | 1 advcl %f %2 612 | 2 mark %f %l 613 | 614 | (S[pt]\NP)/(S[to]\NP) 615 | 2 ((S[pt]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 616 | 1 nsubj %l %f 617 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 618 | 619 | ((S[pt]\NP)/PP)/NP 620 | 3 (((S[pt]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 621 | 1 nsubj %l %f 622 | 2 prep %l %f =PP/NP 623 | 2 xcomp %l %f =PP/(S[ng]\NP) 624 | 2 xcomp %l %f =PP/(S[adj]\NP) 625 | 3 cop %f %l =be 626 | 3 dobj %l %f 627 | 628 | S[asup]\NP 629 | 1 (S[asup]{_}\NP{Y}<1>){_} 630 | 1 dep %f %l # following Stanford parser 631 | 632 | ((N/N)/(N/N))\(S[adj]\NP) 633 | 2 (((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}<1>){_}\(S[adj]{W}<2>\NP{V}){W}){_} 634 | 1 quantmod %f %l # "more than five apples", "more than routine inquiries". Stanford parser does quantmod(five, than), quantmod(routine, than). Possibly right for the former but surely not for the latter! 635 | 2 dep %l %f # following Stanford parser 636 | 637 | ((NP\NP)/S[dcl])\((NP\NP)/NP) 638 | 3 (((NP{Y}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_}\((NP{W}\NP{W}){V}/NP{U}){V}<3>){_} 639 | 1 ref %f %l 640 | 2 rcmodfix %l %f # Bug: should be %1 %f, but not outputting correctly, fixing in python until I can get at it in the code 641 | 3 pobj %f %l 642 | 643 | (S[b]\NP)/S[em] 644 | 2 ((S[b]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 645 | 1 nsubj %l %f 646 | 2 ccomp %l %c =S[em]/S[dcl] # correct, but not sure why these two arg types are singled out 647 | 2 ccomp %l %c =S[em]/S[b] # correct, but ditto 648 | 649 | ((S[b]\NP)/NP)/NP 650 | 3 (((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 651 | 1 nsubj %l %f 652 | 2 iobj %l %f 653 | 3 dobj %l %f 654 | 655 | ((S\NP)/(S\NP))/((S\NP)/(S\NP)) 656 | 1 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}\NP{Z}){Y}){W}<1>){_} 657 | 1 advmod %f %l 658 | 659 | NP/(S[dcl]/NP) 660 | 1 (NP{_}/(S[dcl]{Y}<1>/NP{Z}){Y}){_} 661 | ! (NP{_}/(S[dcl]{Y}<1>/NP{_}){Y}){_} 662 | 1 ignore 663 | # maybe we want the co-indexed NPs for the non-GR case also; check CCGbank at some point 664 | # but keep for now because we want this markedup file (non-! cases) to be consistent with 665 | # the CL-paper 666 | 667 | (S[ng]\NP)/S[em] 668 | 2 ((S[ng]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 669 | 1 nsubj %l %f 670 | 2 ccomp %l %c =S[em]/S[dcl] # correct, but not sure why these two arg types are singled out 671 | 2 ccomp %l %c =S[em]/S[b] # correct, but ditto 672 | 673 | (S[b]\NP)/S[dcl] 674 | 2 ((S[b]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 675 | 1 nsubj %l %f 676 | 2 ccomp %l %f 677 | 678 | (S[b]\NP)/S[qem] 679 | 2 ((S[b]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 680 | 1 nsubj %l %f 681 | 2 advcl %l %c 682 | 683 | ((S\NP)\(S\NP))\NP 684 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_} 685 | 1 mod %f %2 # see "clobbered two years ago" 686 | 2 mod %f %l 687 | 688 | S\S 689 | 1 (S[X]{Y}\S[X]{Y}<1>){_} 690 | 1 mod %f %l 691 | 692 | ((S[ng]\NP)/(S[to]\NP))/NP 693 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 694 | 1 nsubj %l %f 695 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 696 | 3 dobj %l %f # will get processed out by python script 697 | 698 | ((S[b]\NP)/(S[adj]\NP))/NP 699 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 700 | 1 nsubj %l %f 701 | 2 acomp %l %f 702 | 3 dobj %l %f 703 | 704 | (S[pss]\NP)/(S[adj]\NP) 705 | 2 ((S[pss]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 706 | 1 nsubjpass %l %f 707 | 2 acomp %l %f 708 | # "considered vigorous", "made clear" 709 | 710 | (N\N)/N 711 | 2 ((N{Y}\N{Y}<1>){_}/N{Z}<2>){_} 712 | 1 appos %f %2 =bracket 713 | 1 prep %f %l 714 | 2 ignore =bracket 715 | 2 pobj %l %f 716 | 717 | ((S[dcl]\NP)/(S[adj]\NP))/NP 718 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 719 | 1 nsubj %l %f 720 | 2 acomp %l %f 721 | 3 dobj %l %f 722 | 723 | (S[ng]\NP)/(S[adj]\NP) 724 | 2 ((S[ng]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 725 | 1 ignore 726 | 2 cop %f %l =aux 727 | 2 acomp %l %f 728 | # "is being prudent", "is growing weaker" 729 | 730 | (S/S)/(S[ng]\NP) 731 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/(S[ng]{Z}<2>\NP{W}){Z}){_} 732 | 1 prep %f %l 733 | 2 pcomp %l %f 734 | # "After falling for three consecutive months" 735 | # not sure this is always the correct analysis 736 | 737 | (S/S)/N 738 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/N{Z}<2>){_} 739 | 1 mod %f %2 740 | 2 det %f %l # "This year, the index . . ." 741 | 742 | ((S\NP)\(S\NP))/(S[adj]\NP) 743 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[adj]{W}<2>\NP{Z*}){W}){_} 744 | 1 prep %f %l 745 | 2 pobj %l %f 746 | # "play for free", "fell from a year earlier" 747 | 748 | (S[adj]\NP)\(S[adj]\NP) 749 | 1 ((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_} 750 | 1 mod %f %l 751 | 752 | ((S\NP)\(S\NP))\((S\NP)\(S\NP)) 753 | 1 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_} 754 | 1 mod %f %l 755 | 756 | (S[for]/(S[to]\NP))/NP 757 | 2 ((S[for]{_}/(S[to]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 758 | 1 ignore 759 | # As far as I can see, SD does not have a relation between "for" and the verb in the "to" clause 760 | # 1 ccomp _ %l %f # see "for the board to ease interest rates" - B+C have Board as the argument of ccomp, which seems odd 761 | 2 pobj %l %f # I think SD has an odd analysis here, where board is the prepositional object of "for" 762 | 763 | (S[dcl]\NP)/S[qem] 764 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 765 | 1 nsubj %l %f 766 | 2 ccomp %l %f # this seems to be consistent with Stanford parser output even if arg is S[qem]/(S[to]\NP), although I believe that in that case it would be more correct to make it xcomp. Also there are cases where I think it should be advcl %l %c.. 767 | 768 | (N\N)/NP 769 | 2 ((N{Y}\N{Y}<1>){_}/NP{Z}<2>){_} 770 | 1 prep %f %l 771 | 2 pobj %l %f 772 | 773 | ((S[b]\NP)/(S[b]\NP))/NP 774 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 775 | 1 nsubj %l %f 776 | 2 ccomp %l %f 777 | 3 ignore # e.g. "help him avoid being Trumped", SD has no dependency between "help" and "him", probably incorrectly (same for B+C) 778 | 779 | (N/N)\(N/N) 780 | 1 ((N{Y}/N{Y}){Z}\(N{Y}/N{Y}){Z}<1>){_} 781 | 1 advmod %l %f 782 | 783 | (N/N)\(S[adj]\NP) 784 | 2 ((N{Y}/N{Y}<1>){_}\(S[adj]{Z}<2>\NP{W}){Z}){_} 785 | 1 pobj %l %f 786 | 2 mod %f %l 787 | 788 | ((N/N)/(N/N))/((N/N)/(N/N)) 789 | 1 (((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){W}/((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){W}<1>){_} 790 | 1 mod %f %l 791 | 792 | NP/(S[dcl]\NP) 793 | 1 (NP{_}/(S[dcl]{Y}<1>\NP{Z}){Y}){_} 794 | ! (NP{_}/(S[dcl]{Y}<1>\NP{_}){Y}){_} 795 | 1 ignore 796 | # might want to coindex NPs in non-GR case also 797 | 798 | (S[ng]\NP)/S[dcl] 799 | 2 ((S[ng]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 800 | 1 nsubj %l %f 801 | 2 ccomp %l %f 802 | 803 | ((NP\NP)/(S[dcl]\NP))/N 804 | 3 (((NP{Y}\NP{Y}<1>){_}/(S[dcl]{Z}<2>\NP{W}){Z}){_}/N{W}<3>){_} 805 | 1 rcmod %f %2 806 | 2 ignore 807 | 3 poss %f %l # all examples I have looked at are "whose" 808 | # "whose prices go up", "whose activity is affected" 809 | 810 | (S[pt]\NP)/S[em] 811 | 2 ((S[pt]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 812 | 1 nsubj %l %f 813 | 2 ccomp %l %c =S[em]/S[dcl] # correct, but not sure why these two arg types are singled out 814 | 2 ccomp %l %c =S[em]/S[b] # correct, but ditto 815 | 816 | (S/S)\NP 817 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}\NP{Z}<2>){_} 818 | 1 mod %f %2 819 | 2 mod %f %l 820 | 821 | # Inverted aux 822 | (S[q]/(S[b]\NP))/NP 823 | 2 ((S[q]{_}/(S[b]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 824 | ! ((S[q]{Y}/(S[b]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 825 | 1 aux %f %l 826 | 2 nsubjfix %l %f # Problem: should be %1 %f but not being produced, needs post-processing 827 | 828 | ((NP\NP)\(NP\NP))/NP 829 | 2 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}<2>){_}/NP{W}<1>){_} 830 | 1 pobj %l %f 831 | 2 prep %f %l 832 | 833 | PP/(S[adj]\NP) 834 | 1 (PP{_}/(S[adj]{Y}<1>\NP{Z}){Y}){_} 835 | 1 acomp %l %f 836 | 837 | ((N/N)\(N/N))/NP 838 | 2 (((N{Y}/N{Y}){Z}\(N{Y}/N{Y}){Z}<1>){_}/NP{W}<2>){_} 839 | 1 mod %f %l 840 | 2 pobj %l %f 841 | 842 | (S[b]\NP)/(S[b]\NP) 843 | 2 ((S[b]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 844 | 1 ignore =aux 845 | 1 nsubj %l %f 846 | 2 aux %f %l =aux 847 | 2 acomp %l %f 848 | 849 | (S\S)/NP 850 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/NP{Z}<2>){_} 851 | 1 mod %f %l 852 | 2 pobj %l %f 853 | 854 | (NP\NP)\NP 855 | 2 ((NP{Y}\NP{Y}<1>){_}\NP{Z}<2>){_} 856 | 1 mod %f %2 857 | 2 mod %f %l 858 | 859 | ((S[adj]\NP)/(S[adj]\NP))/((S[adj]\NP)/(S[adj]\NP)) 860 | 1 (((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}/((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}<1>){_} 861 | 1 mod %f %l 862 | 863 | ((S[dcl]\NP[expl])/(S[to]\NP))/(S[adj]\NP) 864 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/(S[adj]{V}<3>\NP{U}){V}){_} 865 | 1 nsubj %l %f 866 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 867 | 3 acomp %l %f 868 | 869 | (N/PP)/(S[adj]\NP) 870 | 2 ((N{_}/PP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 871 | 1 mod %2 %f 872 | 2 acomp %l %f 873 | 874 | (NP\NP)\(NP\NP) 875 | 1 ((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}<1>){_} 876 | 1 mod %f %l 877 | 878 | (((S\NP)\(S\NP))/((S\NP)\(S\NP)))/NP 879 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<2>){_}/NP{V}<1>){_} 880 | 1 pobj %l %f 881 | 2 prep %f %l 882 | # "from 8.337 % at last week's sale" 883 | 884 | (((S\NP)\(S\NP))/((S\NP)\(S\NP)))/(((S\NP)\(S\NP))/((S\NP)\(S\NP))) 885 | 1 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}/(((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}<1>){_} 886 | 1 mod %f %l 887 | 888 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))\NP 889 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_}\NP{V}<2>){_} 890 | 1 mod %f %l 891 | 2 mod %f %l 892 | 893 | ((S\NP)\(S\NP))/(S[dcl]\NP) 894 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[dcl]{W}<2>\NP{Z*}){W}){_} 895 | 1 rcmod %f %2 896 | 2 rel %f %l 897 | # "A law went on the books in January that let" 898 | # "An element has been discovered that enhances" 899 | # "It is waiving management fees, which contributes" 900 | # head of rcmod should really be the noun 901 | 902 | (S[pt]\NP)/S[dcl] 903 | 2 ((S[pt]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 904 | 1 nsubj %l %f 905 | 2 ccomp %l %f 906 | 907 | (NP/NP)/(NP/NP) 908 | 1 ((NP{Y}/NP{Y}){Z}/(NP{Y}/NP{Y}){Z}<1>){_} 909 | 1 mod %f %l 910 | 911 | ((S[dcl]\NP)/(S[b]\NP))/NP 912 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 913 | 1 nsubj %l %f 914 | 2 acomp %l %f 915 | 3 ignore 916 | 917 | ((N/N)/(N/N))/(S[asup]\NP) 918 | 2 (((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}<2>){_}/(S[asup]{W}<1>\NP{V}){W}){_} 919 | 1 acomp %l %f 920 | 2 mod %f %l 921 | 922 | S[em]/S[b] 923 | 1 (S[em]{_}/S[b]{Y}<1>){_} 924 | 1 ignore 925 | 926 | ((S[dcl]\NP)/(S[ng]\NP))/NP 927 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 928 | 1 nsubj %l %f 929 | 2 acomp %l %f 930 | 3 dobj %l %f 931 | 932 | (NP\NP)/S[qem] 933 | 2 ((NP{Y}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 934 | 1 advcl %f %2 935 | 2 mod %f %l # no egs in Depbank; assumed to be same as (NP\NP)/S[dcl] 936 | 937 | (NP\NP)/(S[adj]\NP) 938 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 939 | 1 prep %f %l 940 | 2 pobj %l %f 941 | # "more than enough", "sector in general", "increase from a year earlier", etc 942 | # not sure it's right to have an adjective as object of preposition 943 | 944 | (((S\NP)\(S\NP))/(S[ng]\NP))/NP 945 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[ng]{W}<2>\NP{V}){W}){_}/NP{U}<3>){_} 946 | 1 prep %f %l 947 | 2 pcomp %l %f 948 | 3 nsubj %2 %f # doesn't work with %2 in first position 949 | # "with Eli Lilly holding the rest" 950 | 951 | ((N/N)\(N/N))/(N/N) 952 | 2 (((N{Y}/N{Y}){_}\(N{Z}/N{Z}){W}<1>){_}/(N{V}/N{V}){U}<2>){_} 953 | 1 prep %f %l 954 | 2 pobj %l %f # however this should be handled with =bracket as well if I can get it to work 955 | 956 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/S[dcl] 957 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_}/S[dcl]{V}<2>){_} 958 | 1 mod %f %l 959 | 2 mark %f %l # the second "as" in "as soon as the hostages are safe" 960 | 961 | ((S[adj]\NP)\(S[adj]\NP))/S[dcl] 962 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/S[dcl]{W}<2>){_} 963 | 1 rcmod %2 %f 964 | 2 mod %f %l 965 | 966 | ((S[dcl]\NP)/S[em])/NP 967 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/NP{W}<3>){_} 968 | 1 nsubj %l %f 969 | 2 ccomp %l %c =S[em]/S[dcl] 970 | 3 pobj %l %f 971 | 972 | (S[adj]\NP)/S[em] 973 | 2 ((S[adj]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 974 | 1 ignore 975 | 2 ccomp %l %c =S[em]/S[dcl] # correct, but not sure why these two argument types are singled out 976 | 2 ccomp %l %c =S[em]/S[b] # correct, but ditto 977 | 978 | ((S[ng]\NP)/(S[adj]\NP))/NP 979 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 980 | 1 nsubj %l %f 981 | 2 acomp %l %f 982 | 3 dobj %l %f 983 | 984 | ((S[pt]\NP)/(S[to]\NP))/NP 985 | 3 (((S[pt]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 986 | 1 nsubj %l %f 987 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 988 | 3 dobj %l %f # will get processed out by python script 989 | 990 | ((S[ng]\NP)/NP)/NP 991 | 3 (((S[ng]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 992 | 1 nsubj %l %f 993 | 2 iobj %l %f 994 | 3 dobj %l %f 995 | 996 | ((S[dcl]\NP)/NP)/(S[dcl]\NP) 997 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/(S[dcl]{W}<3>\NP{V}){W}){_} 998 | 1 nsubj %l %f 999 | 2 ignore # getting obj through rel pronoun category 1000 | 3 ccomp %l %f 1001 | 1002 | PP/S[qem] 1003 | 1 (PP{_}/S[qem]{Y}<1>){_} 1004 | 1 ccomp %l %f # Stanford parser gives a category pcomp here pcomp(on,arrive) in "depends on how many people arrive", but de Marneffe does not list this category and it is not found in BioInfer development set. Leaving it tentatively as ccomp. 1005 | 1006 | (S[adj]\NP)/S[dcl] 1007 | 2 ((S[adj]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 1008 | 1 ignore 1009 | 2 ccomp %l %f 1010 | 1011 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/N 1012 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_}/N{V}<2>){_} 1013 | 1 mod %f %l 1014 | 2 det %f %l 1015 | 1016 | S[dcl]/NP 1017 | 1 (S[dcl]{_}/NP{Y}<1>){_} 1018 | 1 dobj %l %f 1019 | 1020 | (NP\NP)/(S[pss]\NP) 1021 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{W}){Z}){_} 1022 | 1 advcl %f %2 1023 | 2 mark %f %l 1024 | # "as calculated", "than previously reported", sometimes also used for LRB 1025 | 1026 | (S/S)/(S[adj]\NP) 1027 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 1028 | 1 mod %f %l 1029 | 2 acomp %l %f 1030 | 1031 | S[qem]/(S[to]\NP) 1032 | 1 (S[qem]{_}/(S[to]{Y}<1>\NP{Z}){Y}){_} 1033 | 1 advmod %k %f # actually this needs to vary quite a bit. "wondering who this man is": dobj(is,who). "wondering whether to live in London": complm(live,whether). "learning how to read": advmod(read,how). 1034 | 1035 | ((S[pss]\NP)/(S[to]\NP))/PP 1036 | 3 (((S[pss]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{V}<3>){_} 1037 | 1 nsubjpass %l %f 1038 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 1039 | 3 prep %l %f =PP/NP 1040 | 1041 | ((S[dcl]\NP[expl])/S[em])/(S[adj]\NP) 1042 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/S[em]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1043 | 1 nsubj %l %f 1044 | 2 ccomp %3 %c =S[em]/S[dcl] # "It is unnecessary that we win." 1045 | 3 cop %f %l 1046 | 1047 | (S[pss]\NP)/S[em] 1048 | 2 ((S[pss]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 1049 | 1 nsubjpass %l %f 1050 | 2 ccomp %l %c =S[em]/S[dcl] # correct, but not sure why these two arg types are singled out 1051 | 2 ccomp %l %c =S[em]/S[b] # correct, but ditto 1052 | 1053 | ((S[dcl]\NP)/PP)/(S[adj]\NP) 1054 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1055 | 1 nsubj %l %f 1056 | 2 prep %l %f =PP/NP 1057 | 3 acomp %l %f 1058 | 1059 | ((NP\NP)\(NP\NP))/((NP\NP)\(NP\NP)) 1060 | 1 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}/((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}<1>){_} 1061 | 1 mod %f %l 1062 | 1063 | PP\PP 1064 | 1 (PP{Y}\PP{Y}<1>){_} 1065 | 1 mod %f %l 1066 | 1067 | (S[dcl]\NP[expl])/(S[b]\NP) 1068 | 2 ((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[b]{Z}<2>\NP{W}){Z}){_} 1069 | ! ((S[dcl]{Z}\NP[expl]{Y}<1>){Z}/(S[b]{Z}<2>\NP{Y}){Z}){_} 1070 | 1 ignore =aux 1071 | 1 nsubj %l %f 1072 | 2 aux %f %l =aux 1073 | 2 acomp %l %f 1074 | 1075 | ((S[b]\NP)/(S[ng]\NP))/NP 1076 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1077 | 1 nsubj %l %f 1078 | 2 acomp %l %f 1079 | 3 dobj %l %f 1080 | 1081 | (PP\PP)\NP 1082 | 2 ((PP{Y}\PP{Y}<1>){_}\NP{Z}<2>){_} 1083 | 1 mod %f %2 # not consistent with Depbank but difficult to get right 1084 | 2 mod %f %l 1085 | 1086 | (S[dcl]\PP)/NP 1087 | 2 ((S[dcl]{_}\PP{Y}<1>){_}/NP{Z}<2>){_} 1088 | 1 nsubj %l %f # PPs can be subjects according to Ted's tech report (but no examples in Depbank) 1089 | 2 dobj %l %f 1090 | 1091 | S[dcl]/S[dcl] 1092 | 1 (S[dcl]{_}/S[dcl]{Y}<1>){_} 1093 | 1 complm %f %l 1094 | 1095 | # initial "how" or "why" 1096 | S[wq]/S[q] 1097 | 1 (S[wq]{_}/S[q]{Y}<1>){_} 1098 | 1 advmod %f %l 1099 | 1100 | # adverbial-like determiner, e.g. "what day" 1101 | (S[wq]/S[q])/N 1102 | 2 ((S[wq]{_}/S[q]{Y}<1>){_}/N{Z}<2>) 1103 | 1 mod %f %l # ? 1104 | 2 det %f %l 1105 | 1106 | S/S[dcl] 1107 | 1 (S{_}/S[dcl]{Y}<1>){_} 1108 | 1 complm %f %l 1109 | 1110 | (S[dcl]\(S[adj]\NP))/NP 1111 | 2 ((S[dcl]{_}\(S[adj]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 1112 | 1 subj %l %f 1113 | 2 acomp %l %f =be 1114 | 2 dobj %l %f 1115 | 1116 | ((S\NP)\(S\NP))/(S[pss]\NP) 1117 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[pss]{W}<2>\NP{Z*}){W}){_} 1118 | 1 advcl %f %2 1119 | 2 mark %f %l 1120 | # "when delivered", "as measured", "when overexpressed" 1121 | 1122 | ((S\NP)/(S\NP))/N 1123 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<2>\NP{Z}){Y}){_}/N{W}<1>){_} 1124 | 1 det %f %l 1125 | 2 mod %f %1 1126 | 1127 | ((NP\NP)/(S[dcl]\NP))\(NP/NP) 1128 | 3 (((NP{Y}\NP{Y}<1>){_}/(S[dcl]{Z}<2>\NP{W}){Z}){_}\(NP{V}/NP{V}){U}<3>){_} 1129 | 1 ref %f %l 1130 | 2 rcmod %1 %f # broken: doesn't like %1 in first slot 1131 | 3 rel %2 %f # broken: doesn't like %2 in first slot 1132 | # "all of which", "several of which fell" 1133 | # Per BioInfer, should also have relations pobj of which, prep several of 1134 | 1135 | (((NP\NP)/(NP\NP))\((NP\NP)/(NP\NP)))/((NP\NP)/(NP\NP)) 1136 | 2 ((((NP{Y}\NP{Y}){Z}/(NP{Y}\NP{Y}){Z}){_}\((NP{Y}\NP{Y}){Z}/(NP{Y}\NP{Y}){Z}){W}<1>){_}/((NP{Y}\NP{Y}){U}/(NP{Y}\NP{Y}){U}){V}<2>){_} 1137 | 1 pobj %l %f # no egs but similar to ((N/N)\(N/N))/(N/N) 1138 | 2 mod %l %f 1139 | 1140 | ((NP\NP)/(NP\NP))/NP 1141 | 3 (((NP{Y}\NP{Y}<1>){_}/(NP{Z}\NP{Z}){W}<3>){_}/NP{V}<2>){_} 1142 | 1 mod %f %l 1143 | 2 pobj %l %f 1144 | 3 ignore 1145 | # "from the fall of 1962 through ..." difficult to get looking like B+C 1146 | 1147 | # Subject wh-pronoun 1148 | S[wq]/(S[dcl]\NP) 1149 | 1 (S[wq]{_}/(S[dcl]{Y}<1>\NP{Z}){Y}){_} 1150 | 1 nsubj %f %l 1151 | 1152 | NP/S[dcl] 1153 | 1 (NP{_}/S[dcl]{Y}<1>){_} 1154 | 1 dobj %f %l # a bit difficult - "This is what the government calls a disaster" - the CCGbank analysis is odd so this may be correct but the verb is going to mistakenly have two dobjs - although this seems to match Stanford parser output! 1155 | 1156 | (PP\PP)/NP 1157 | 2 ((PP{Y}\PP{Y}<1>){_}/NP{Z}<2>){_} 1158 | 1 prep %f %l 1159 | 2 pobj %l %f 1160 | 1161 | ((S[dcl]\NP)/S[dcl])/NP 1162 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_}/NP{W}<3>){_} 1163 | 1 nsubj %l %f 1164 | 2 ccomp %l %f 1165 | 3 dobj %l %f 1166 | 1167 | ((S\NP)/(S\NP))/S[dcl] 1168 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/S[dcl]{W}<2>){_} 1169 | 1 advcl %f %2 # done, but need to test. Example sentence "If you're wondering, the debt is large." 1170 | 2 mark %f %l 1171 | 1172 | (S[adj]\NP)/((S[to]\NP)/NP) 1173 | 2 ((S[adj]{_}\NP{Y}<1>){_}/((S[to]{Z}<2>\NP{W}){Z}/NP{V}){Z}){_} 1174 | 1 ignore 1175 | 2 acomp %l %f 1176 | 1177 | ((S[adj]\NP)/PP)/NP 1178 | 3 (((S[adj]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 1179 | 1 ignore 1180 | 2 mod %3 %f # "up .. from .." construction 1181 | 3 dobj %l %f 1182 | 1183 | (((S\NP)\(S\NP))/PP)/NP 1184 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/PP{W}<2>){_}/NP{V}<3>){_} 1185 | 1 mod %f %l 1186 | 2 mod %3 %f # "up .. from .." construction 1187 | 3 dobj %l %f 1188 | 1189 | N/S[dcl] 1190 | 1 (N{_}/S[dcl]{Y}<1>){_} 1191 | 1 rcmod %l %f # "the fact they installed circuit breakers" - there are linguistic analyses (not sure of references) of complements to nouns like "fact" as being relative clauses; this would be consistent. Haven't been able to find an example in BioInfer but perhaps it will come up during further testing. 1192 | 1193 | (((S[b]\NP)/PP)/PP)/NP 1194 | 4 ((((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_}/NP{V}<4>){_} 1195 | 1 nsubj %l %f 1196 | 2 mod %f %l # don't have two iobj's 1197 | 3 prep %l %f =PP/NP 1198 | 4 dobj %l %f 1199 | 1200 | ((S[adj]\NP)/S[em])/(S[adj]\NP) 1201 | 3 (((S[adj]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1202 | 1 ignore 1203 | 2 ccomp %3 %c =S[em]/S[dcl] # "so weak that ..." construction 1204 | 3 advmod %f %l 1205 | 1206 | ((NP\NP)/PP)/NP 1207 | 3 (((NP{Y}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 1208 | 1 mod %f %l 1209 | 2 mod %3 %f # no egs but this is the "up .. from .." construction 1210 | 3 pobj %l %f 1211 | 1212 | ((S\NP)/(S\NP))/PP 1213 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/PP{W}<2>){_} 1214 | # 1 (pmod %f %l) POS==IN 1215 | 1 mod %f %l # no egs so follow ((S\NP)\(S\NP))/PP 1216 | 2 dep %l %f =PP/NP # to match Stanford parser output 1217 | # do we want to support POS constraints (and pmod)? 1218 | 1219 | (S[ng]\NP)/(S[pt]\NP) 1220 | 2 ((S[ng]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 1221 | ! ((S[ng]{Z}\NP{Y}<1>){Z}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 1222 | 1 ignore =aux 1223 | 1 nsubj %l %f 1224 | 2 aux %f %l =aux 1225 | 2 acomp %l %f 1226 | 1227 | (PP/PP)/NP 1228 | 2 ((PP{_}/PP{Y}<1>){_}/NP{Z}<2>){_} 1229 | 1 mod %2 %f 1230 | 2 pobj %l %f 1231 | 1232 | (S[ng]\NP)/S[qem] 1233 | 2 ((S[ng]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 1234 | 1 nsubj %l %f 1235 | 2 ccomp %l %c # this seems to be consistent with Stanford parser output even if arg is S[qem]/(S[to]\NP), although I believe that in that case it would be more correct to make it xcomp. Also there are cases where I think it should be advcl. 1236 | 1237 | S[inv]/NP 1238 | 1 (S[inv]{_}/NP{Y}<1>){_} 1239 | 1 acomp %l %f =be 1240 | 1 dobj %l %f # technically this should be nsubj "I attended, and so did Mark." Stanford parser gets this right if there's no conjunction, e.g. if the sentence is "So says the Commonwealth Fund". But a conjunction confuses it into thinking that "attended" and "did" are conjoined, with "Mark" the dobj of "did". I don't think we can allow for this in the markedup file, just something to be aware of. 1241 | 1242 | ((S[pt]\NP)/NP)/NP 1243 | 3 (((S[pt]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 1244 | 1 nsubj %l %f 1245 | 2 dobj %l %f 1246 | 3 iobj %l %f 1247 | 1248 | ((S[dcl]\NP)/PP)/PP 1249 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_} 1250 | 1 nsubj %l %f 1251 | 2 prep %l %f =PP/NP 1252 | 2 xcomp %l %c =PP/(S[to]\NP) 1253 | 3 prep %l %f =PP/NP 1254 | 3 xcomp %l %c =PP/(S[to]\NP) 1255 | 1256 | (N/N)/N 1257 | 2 ((N{Y}/N{Y}<1>){_}/N{Z}<2>){_} 1258 | 1 appos %f %2 # Normally a left bracket. The Stanford parser distinguishes abbreviations, but BioInfer merges abbrevs into the appos category. 1259 | 2 det %f %l =fulldet # second "the" in "the Friday the 13th plunge" 1260 | 2 pobj %l %f 1261 | # another way would be to make this prep/pobj and post-process it 1262 | 1263 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/(((S\NP)\(S\NP))\((S\NP)\(S\NP))) 1264 | 1 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}/(((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}<1>){_} 1265 | 1 mod %f %l # this could be num, amod, or tmod (at least). Will need to do some postprocessing, at least for the numbers. 1266 | 1267 | ((S[ng]\NP)/(S[b]\NP))/NP 1268 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1269 | 1 nsubj %l %f 1270 | 2 ccomp %l %f 1271 | 3 ignore 1272 | 1273 | (NP\NP)/(S[b]\NP) 1274 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{W}){Z}){_} 1275 | 1 rcmod %f %2 1276 | 2 nsubj %f %l 1277 | 1278 | ((S/S)/(S[ng]\NP))/NP 1279 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/(S[ng]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1280 | 1 prep %f %l 1281 | 2 partmod %3 %f 1282 | 3 ignore 1283 | 1284 | (S[dcl]\NP)/(S[dcl]\NP) 1285 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/(S[dcl]{Z}<2>\NP{W}){Z}){_} 1286 | 1 nsubj %l %f 1287 | 2 ccomp %l %f # some odd analysis in CCGbank 1288 | 1289 | (S[qem]/S[dcl])/(S[adj]\NP) 1290 | 2 ((S[qem]{_}/S[dcl]{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 1291 | 1 advmod %f %l # the lexical item is (pretty much?) always "how" 1292 | 2 advmod %l %f # this seems to be consistent with the parser, though it doesn't seem correct to me "how far you go", "how long we should wait" 1293 | 1294 | (S[dcl]\NP)/S[for] 1295 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_} 1296 | 1 nsubj %l %f 1297 | 2 ccomp %l %c 1298 | 1299 | ((S[b]\NP)/NP)/(S[adj]\NP) 1300 | 3 (((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1301 | 1 nsubj %l %f 1302 | 2 dobj %l %f 1303 | 3 acomp %l %f 1304 | 1305 | ((S[dcl]\NP)/NP)/PP 1306 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/PP{W}<3>){_} 1307 | 1 nsubj %l %f 1308 | 2 dobj %l %f 1309 | 3 prep %l %f =PP/NP 1310 | 1311 | (S/S)/(S[pss]\NP) 1312 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/(S[pss]{Z}<2>\NP{W}){Z}){_} 1313 | 1 advcl %f %2 # this is not outputting for some reason 1314 | 2 advmod %f %l 1315 | 1316 | ((S[adj]\NP)\(S[adj]\NP))/PP 1317 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/PP{W}<2>){_} 1318 | 1 advmod %f %l 1319 | 2 prep %l %f 1320 | # "more than in 1981", "puttable back to the company", "available free of charge" 1321 | 1322 | (PP\NP)/NP 1323 | 2 ((PP{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 1324 | 1 mod %f %l 1325 | 2 pobj %l %f 1326 | 1327 | (S[qem]/(S[dcl]/NP))/N 1328 | 2 ((S[qem]{_}/(S[dcl]{Y}<1>/NP{Z}){Y}){_}/N{W}<2>){_} 1329 | 1 ccomp %l %f 1330 | 2 dobj %l %f # no egs 1331 | 1332 | (((S\NP)\(S\NP))\NP)/NP 1333 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_}/NP{V}<3>){_} 1334 | 1 mod %f %2 # no egs; guessing what B+C would do 1335 | 2 mod %f %l 1336 | 3 pobj %l %f 1337 | 1338 | (PP\PP)/(PP\PP) 1339 | 1 ((PP{Y}\PP{Y}){Z}/(PP{Y}\PP{Y}){Z}<1>){_} 1340 | 1 mod %f %l 1341 | 1342 | (S[dcl]/S[dcl])/NP 1343 | 2 ((S[dcl]{_}/S[dcl]{Y}<1>){_}/NP{Z}<2>){_} 1344 | 1 ccomp %l %f 1345 | 2 nsubjpass %l %f 1346 | 1347 | (PP/(S[ng]\NP))/NP 1348 | 2 ((PP{_}/(S[ng]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 1349 | 1 ccomp %l %f 1350 | 2 ignore # obj becomes subj of verb through co-indexing 1351 | 1352 | S[qem]/(S[dcl]\NP) 1353 | 1 (S[qem]{_}/(S[dcl]{Y}<1>\NP{Z}){Y}){_} 1354 | ! (S[qem]{_}/(S[dcl]{Y}<1>\NP{_}){Y}){_} 1355 | 1 ignore 1356 | # might want to get co-indexed NPs on non-GR case also 1357 | 1358 | S[poss]/S[dcl] 1359 | 1 (S[poss]{_}/S[dcl]{Y}<1>){_} 1360 | 1 ccomp %l %f 1361 | 1362 | ((S\NP)\(S\NP))/S[em] 1363 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S[em]{W}<2>){_} 1364 | 1 advcl %f %l 1365 | 2 ccomp %l %c =S[em]/S[dcl] 1366 | 1367 | (N/N)/(S[asup]\NP) 1368 | 2 ((N{Y}/N{Y}<1>){_}/(S[asup]{Z}<2>\NP{W}){Z}){_} 1369 | 1 mod %f %l 1370 | 2 acomp %l %f 1371 | 1372 | conj\conj 1373 | 1 (conj{Y}\conj{Y}<1>){_} 1374 | 1 mod %f %l 1375 | 1376 | ((NP\NP)/(S[ng]\NP))/NP 1377 | 3 (((NP{Y}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1378 | 1 mod %f %l 1379 | 2 ccomp %l %f 1380 | 3 ignore # obj becomes subj of verb through co-indexing 1381 | 1382 | ((S[adj]\NP)\(S[adj]\NP))/(S[adj]\NP) 1383 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/(S[adj]{W}<2>\NP{V}){W}){_} 1384 | 1 mod %f %l 1385 | 2 acomp %l %f 1386 | 1387 | ((S[dcl]\NP[expl])/(NP\NP))/NP 1388 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/(NP{Z}\NP{Z}){W}<2>){_}/NP{V}<3>){_} 1389 | 1 nsubj %l %f 1390 | 2 mod %l %f # odd analysis in CCGbank for this category 1391 | 3 dobj %l %f 1392 | 1393 | (NP/NP)/NP 1394 | 2 ((NP{Y}/NP{Y}<1>){_}/NP{Z}<2>){_} 1395 | 1 mod %f %l 1396 | 2 pobj %l %f 1397 | 1398 | ((S[pt]\NP)/(S[adj]\NP))/NP 1399 | 3 (((S[pt]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1400 | 1 nsubj %l %f 1401 | 2 acomp %l %f 1402 | 3 dobj %l %f 1403 | 1404 | ((S\NP)\(S\NP))/(S[to]\NP) 1405 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[to]{W}<2>\NP{V}){W}){_} 1406 | 1 mod %f %l 1407 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 1408 | 1409 | (S[dcl]\NP)/S 1410 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S{Z}<2>){_} 1411 | 1 nsubj %l %f 1412 | 2 ccomp %l %f 1413 | 1414 | (S/S)\(S/S) 1415 | 1 ((S[X]{Y}/S[X]{Y}){Z}\(S[X]{Y}/S[X]{Y}){Z}<1>){_} 1416 | 1 mod %f %l 1417 | 1418 | (((S\NP)\(S\NP))\NP)/S[dcl] 1419 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_}/S[dcl]{V}<3>){_} 1420 | 1 mod %f %2 # guess at what B+C would have 1421 | 2 mod %f %l 1422 | 3 ccomp %l %f 1423 | 1424 | ((S[adj]\NP)/(S[to]\NP))/(S[adj]\NP) 1425 | 2 (((S[adj]{Y}\NP{Z}){Y}/(S[to]{W}<2>\NP{Z}){W}){_}/(S[adj]{Y}<1>\NP{Z}){Y}){_} 1426 | 1 mod %f %l 1427 | 2 xcomp %1 %k =(S[to]\NP)/(S[b]\NP) # guess at what B+C would have "too difficult to maintain" 1428 | 1429 | (N/N)/NP 1430 | 2 ((N{Y}/N{Y}<1>){_}/NP{Z}<2>){_} 1431 | 1 prep %f %l 1432 | 2 pobj %l %f 1433 | # "little, if any, potentially lingering damage"; also used for LRB most of the time 1434 | 1435 | (S[ng]\NP)/(S[ng]\NP) 1436 | 2 ((S[ng]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 1437 | 1 nsubj %l %f 1438 | 2 acomp %l %f 1439 | 1440 | (((S[dcl]\NP)/PP)/PP)/NP 1441 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_}/NP{V}<4>){_} 1442 | 1 nsubj %l %f 1443 | 2 mod %f %l # don't have two iobj's 1444 | 3 prep %l %f =PP/NP 1445 | 4 dobj %l %f 1446 | 1447 | ((S[dcl]\NP)/NP)/(S[adj]\NP) 1448 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1449 | 1 nsubj %l %f 1450 | 2 dobj %l %f 1451 | 3 acomp %l %f 1452 | 1453 | ((S[b]\NP)/NP)/PP 1454 | 3 (((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/PP{W}<3>){_} 1455 | 1 nsubj %l %f 1456 | 2 dobj %l %f 1457 | 3 prep %l %f =PP/NP 1458 | 1459 | S[qem]/(S[dcl]/NP) 1460 | 1 (S[qem]{_}/(S[dcl]{Y}<1>/NP{Z}){Y}){_} 1461 | ! (S[qem]{_}/(S[dcl]{Y}<1>/NP{_}){Y}){_} 1462 | 1 ccomp %l %f 1463 | # might want to coindex object in non-GR case also 1464 | 1465 | (S[pss]\NP)/(PP/NP) 1466 | 2 ((S[pss]{_}\NP{Y}<1>){_}/(PP{Z}<2>/NP{W}){Z}){_} 1467 | 1 nsubjpass %l %f 1468 | 2 prep %l %f 1469 | 1470 | (((S\NP)\(S\NP))/(S[to]\NP))/N 1471 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[to]{W}<2>\NP{V}){W}){_}/N{U}<3>){_} 1472 | 1 purpcl %f %2 # needs post-processing in python 1473 | 2 mark %k %l =(S[to]\NP)/(S[b]\NP) # "in order to become" 1474 | 3 dep %1 %l # needs post-processing in python 1475 | 1476 | (S[dcl]\(S[ng]\NP))/NP 1477 | 2 ((S[dcl]{_}\(S[ng]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 1478 | 1 acomp %l %f 1479 | 2 nsubjpass %l %f 1480 | # might want to coindex NPs 1481 | 1482 | ((S[dcl]\NP[expl])/S[qem])/(S[adj]\NP) 1483 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/S[qem]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1484 | 1 nsubj %l %f 1485 | 2 ccomp %l %f 1486 | 3 acomp %l %f 1487 | 1488 | ((S\NP)/(S\NP))/(S[adj]\NP) 1489 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[adj]{W}<2>\NP{Z*}){W}){_} 1490 | 1 mod %f %l 1491 | 2 acomp %l %f # consistent with ((S\NP)\(S\NP))/(S[adj]\NP) 1492 | 1493 | ((S[dcl]\NP[expl])/(S[to]\NP))/NP 1494 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/NP{V}<3>){_} 1495 | 1 nsubj %l %f 1496 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 1497 | 3 dobj %l %f # will get processed out by python script 1498 | 1499 | ((S[b]\NP)/PP)/(S[adj]\NP) 1500 | 3 (((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1501 | 1 nsubj %l %f 1502 | 2 prep %l %f =PP/NP 1503 | 3 acomp %l %f 1504 | 1505 | ((NP/NP)/(NP/NP))\(S[adj]\NP) 1506 | 2 (((NP{Y}/NP{Y}){Z}/(NP{Y}/NP{Y}){Z}<1>){_}\(S[adj]{W}<2>\NP{V}){W}){_} 1507 | 1 mod %f %l # another "less than" construction 1508 | 2 mod %f %l 1509 | 1510 | (S[qem]/(S[dcl]/(S[adj]\NP)))/(S[adj]\NP) 1511 | 2 ((S[qem]{_}/(S[dcl]{Y}<1>/(S[adj]{Z}\NP{W}){Z}){Y}){_}/(S[adj]{Z}<2>\NP{V}){Z}){_} 1512 | 1 mod %f %l 1513 | 2 dobj %l %f # "how big it is" consistent with B+C 1514 | 1515 | ((S\NP)/(S\NP))/(S[ng]\NP) 1516 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[ng]{W}<2>\NP{Z*}){W}){_} 1517 | 1 mod %f %l 1518 | 2 acomp %l %f 1519 | 1520 | ((S\NP)/(S\NP))\(S[adj]\NP) 1521 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}\(S[adj]{W}<2>\NP{V}){W}){_} 1522 | 1 mod %f %l # another "less than" construction 1523 | 2 mod %f %l 1524 | 1525 | ((S[adj]\NP)\(S[adj]\NP))/(S[ng]\NP) 1526 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/(S[ng]{W}<2>\NP{V}){W}){_} 1527 | 1 mod %f %l # "reponsible for buying" 1528 | 2 acomp %l %f 1529 | 1530 | ((S\NP)\(S\NP))/S[qem] 1531 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S[qem]{W}<2>){_} 1532 | 1 mod %f %l 1533 | 2 ccomp %l %f # consistent with ((S\NP)\(S\NP))/S[dcl] 1534 | 1535 | ((S[adj]\NP)\(S[adj]\NP))/((S[adj]\NP)\(S[adj]\NP)) 1536 | 1 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}\NP{Z}){Y}){W}/((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}\NP{Z}){Y}){W}<1>){_} 1537 | 1 advmod %f %l 1538 | 1539 | (S[dcl]/(S[b]\NP))/NP 1540 | 2 ((S[dcl]{_}/(S[b]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 1541 | ! ((S[dcl]{Y}/(S[b]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 1542 | 1 aux %f %l 1543 | 2 ignore # ? . consistent with (S[q]/(S[b]\NP))/NP 1544 | 1545 | S[to]\NP 1546 | 1 (S[to]{_}\NP{Y}<1>){_} 1547 | 1 ignore 1548 | 1549 | # Inverted aux 1550 | (S[q]/NP)/NP 1551 | 2 ((S[q]{_}/NP{Y}<1>){_}/NP{Z}<2>){_} 1552 | 1 cop %f %l =be 1553 | 1 dobj %l %f 1554 | 2 nsubjfix %l %f =be # Problem: should be %1 %f, but not being produced; post-process 1555 | 2 nsubj %l %f 1556 | 1557 | # Inverted aux where the underlying subject is a sentence (for quotations, specifically "whose slogan is live free or die") 1558 | (S[q]/NP)/(S[b]\NP) 1559 | 2 ((S[q]{_}/NP{Y}<1>){_}/(S[b]{Z}<2>\NP){Z}){_} 1560 | 1 cop %f %l =be 1561 | 1 dobj %l %f 1562 | 2 csubjfix %l %f =be # Problem: should be %1 %f, but not being produced; post-process 1563 | 2 csubj %l %f 1564 | 1565 | # Inverted aux where the underlying subject is a sentence (for quotations, specifically "whose line is can we talk") 1566 | (S[q]/NP)/S[q] 1567 | 2 ((S[q]{_}/NP{Y}<1>){_}/S[q]{Z}<2>){_} 1568 | 1 cop %f %l =be 1569 | 1 dobj %l %f 1570 | 2 csubjfix %l %f =be # Problem: should be %1 %f, but not being produced; post-process 1571 | 2 csubj %l %f 1572 | 1573 | # Inverted aux where the underlying subject is a sentence (for quotations, specifically "whose slogan is quality is job 1") 1574 | (S[q]/NP)/S[dcl] 1575 | 2 ((S[q]{_}/NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 1576 | 1 cop %f %l =be 1577 | 1 dobj %l %f 1578 | 2 csubjfix %l %f =be # Problem: should be %1 %f, but not being produced; post-process 1579 | 2 csubj %l %f 1580 | 1581 | # Inverted aux with expletive 1582 | (S[q]/NP)/NP[thr] 1583 | 2 ((S[q]{_}/NP{Y}<1>){_}/NP[thr]{Z}<2>){_} 1584 | 1 nsubj %l %f =be # ? 1585 | 1 dobj %l %f # ? 1586 | 2 expl %l %f # ? 1587 | 1588 | ((S[dcl]\S[dcl])\NP)/NP 1589 | 3 (((S[dcl]{_}\S[dcl]{Y}<1>){_}\NP{Z}<2>){_}/NP{W}<3>){_} 1590 | 1 advcl %l %f 1591 | 2 nsubj %l %f 1592 | 3 dobj %l %f 1593 | 1594 | ((NP\NP)/(S[to]\NP))/NP 1595 | 3 (((NP{Y}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1596 | 1 mod %f %l 1597 | 2 infmod %3 %k =(S[to]\NP)/(S[b]\NP) # "authority for administration to raise" 1598 | 3 pobj %l %f 1599 | 1600 | (S/S)/(S[dcl]\NP) 1601 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/(S[dcl]{Z}<2>\NP{W}){Z}){_} 1602 | 1 advcl %f %2 1603 | 2 ignore # "What's more ..." 1604 | # co-index subj with lexical item? (so what becomes subj of 's) 1605 | 1606 | ((S[b]\NP)/PP)/PP 1607 | 3 (((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_} 1608 | 1 nsubj %l %f 1609 | 2 mod %f %l # don't have two iobj's 1610 | 3 prep %l %f =PP/NP 1611 | 1612 | (N/N)/(S[adj]\NP) 1613 | 2 ((N{Y}/N{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 1614 | 1 amod %f %2 1615 | 2 advmod %f %l 1616 | # "a touching, if self-congratulatory, ad", "this too private garden", 1617 | # also used for LRB "the (quarterly) refunding" 1618 | 1619 | (((S\NP)\(S\NP))/((S\NP)\(S\NP)))/(S[asup]\NP) 1620 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<2>){_}/(S[asup]{V}<1>\NP{U}){V}){_} 1621 | 1 dep %l %f # following Stanford parser 1622 | 2 advmod %f %l # not sure this is correct 1623 | # "at least through the first quarter", "at least partially" 1624 | 1625 | ((S\NP)\(S\NP))/(S[dcl]/NP) 1626 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[dcl]{W}<2>/NP{V}){W}){_} 1627 | 1 rcmod %f %2 1628 | 2 ignore # consistent with ((S\NP)\(S\NP))/(S[dcl]\NP) 1629 | 1630 | (NP/NP)/(S[asup]\NP) 1631 | 2 ((NP{Y}/NP{Y}<1>){_}/(S[asup]{Z}<2>\NP{W}){Z}){_} 1632 | 1 advmod %f %l 1633 | 2 dep %l %f 1634 | # "at least a pleasant bonus" 1635 | 1636 | ((NP\NP)/(NP\NP))/((NP\NP)/(NP\NP)) 1637 | 1 (((NP{Y}\NP{Y}){Z}/(NP{Y}\NP{Y}){Z}){W}/((NP{Y}\NP{Y}){Z}/(NP{Y}\NP{Y}){Z}){W}<1>){_} 1638 | 1 mod %f %l 1639 | 1640 | ((N/N)\(N/N))/N 1641 | 2 (((N{Y}/N{Y}){Z}\(N{Y}/N{Y}){Z}<2>){_}/N{W}<1>){_} 1642 | 1 pobj %l %f 1643 | 2 mod %f %l 1644 | 1645 | ((S[dcl]\NP)/(S[pss]\NP))/NP 1646 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1647 | 1 nsubj %l %f 1648 | 2 acomp %l %f 1649 | 3 dobj %l %f 1650 | 1651 | ((S[dcl]\NP[expl])/S[for])/(S[adj]\NP) 1652 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/S[for]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 1653 | 1 nsubj %l %f 1654 | 2 ccomp %3 %f # see "It's terrific for advertisers ..." 1655 | 3 acomp %l %f 1656 | 1657 | (NP/NP)\(S[adj]\NP) 1658 | 2 ((NP{Y}/NP{Y}<1>){_}\(S[adj]{Z}<2>\NP{W}){Z}){_} 1659 | 1 mod %f %l # another "less than" construction 1660 | 2 mod %f %l 1661 | 1662 | # Object wh-pronoun 1663 | S[wq]/(S[q]/NP) 1664 | 1 (S[wq]{_}/(S[q]{Y}<1>/NP{Z}){Y}){_} 1665 | ! (S[wq]{_}/(S[q]{Y}<1>/NP{_}){Y}){_} 1666 | 1 ignore 1667 | # might want the co-indexing in the non-GR case also 1668 | 1669 | ((S/S)\(S/S))/S[dcl] 1670 | 2 (((S[X]{Y}/S[X]{Y}){Z}\(S[X]{Y}/S[X]{Y}){Z}<2>){_}/S[dcl]{W}<1>){_} 1671 | 2 mod %f %l 1672 | 1 ccomp %l %f 1673 | 1674 | ((S/S)\(S/S))/NP 1675 | 2 (((S[X]{Y}/S[X]{Y}){Z}\(S[X]{Y}/S[X]{Y}){Z}<2>){_}/NP{W}<1>){_} 1676 | 1 pobj %l %f 1677 | 2 mod %l %f 1678 | 1679 | (S[qem]/(S[dcl]\NP))/N 1680 | 2 ((S[qem]{_}/(S[dcl]{Y}<1>\NP{Z*}){Y}){_}/N{Z}<2>){_} 1681 | 1 ccomp %l %f 1682 | 2 ignore 1683 | 1684 | ((S[pss]\NP)/PP)/PP 1685 | 3 (((S[pss]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_} 1686 | 1 nsubjpass %l %f 1687 | 2 mod %f %l # don't have two iobj's 1688 | 3 prep %l %f =PP/NP 1689 | 1690 | ((S\NP)/(S\NP))/(S[pss]\NP) 1691 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[pss]{W}<2>\NP{V}){W}){_} 1692 | 1 mod %f %l 1693 | 2 acomp %l %f 1694 | 1695 | ((S\NP)\(S\NP))/S[inv] 1696 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S[inv]{W}<2>){_} 1697 | 1 advcl %f %l 1698 | 2 ccomp %l %f 1699 | 1700 | (((S\NP)\(S\NP))/(S[b]\NP))/NP 1701 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[b]{W}<2>\NP{V*}){W}){_}/NP{V}<3>){_} 1702 | 1 mod %l %f 1703 | 2 ccomp %l %f 1704 | 3 ignore 1705 | 1706 | ((S\NP)\(S\NP))/(S[b]\NP) 1707 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[b]{W}<2>\NP{Z*}){W}){_} 1708 | 1 mod %f %l 1709 | 2 acomp %l %f 1710 | 1711 | ((S[ng]\NP)/(S[ng]\NP))/NP 1712 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_}/NP{W}<3>){_} 1713 | 1 nsubj %l %f 1714 | 2 acomp %l %f 1715 | 3 pobj %l %f 1716 | # the co-indexing looks wrong here 1717 | 1718 | ((S[ng]\NP)/PP)/PP 1719 | 3 (((S[ng]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_} 1720 | 1 nsubj %l %f 1721 | 2 mod %f %l # don't have two iobj's 1722 | 3 prep %l %f =PP/NP 1723 | 1724 | ((S[dcl]\NP)/(S[to]\NP))/PP 1725 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{V}<3>){_} 1726 | 1 nsubj %l %f 1727 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 1728 | 3 prep %l %f =PP/NP 1729 | 1730 | ((S[adj]\NP)/PP)/PP 1731 | 3 (((S[adj]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_} 1732 | 1 ignore 1733 | 2 mod %f %l # don't have two iobj's 1734 | 3 prep %l %f =PP/NP 1735 | 1736 | (S\S)/(S\S) 1737 | 1 ((S[X]{Y}\S[X]{Y}){Z}/(S[X]{Y}\S[X]{Y}){Z}<1>){_} 1738 | 1 mod %f %l 1739 | 1740 | (((S\NP)\(S\NP))/(S[to]\NP))/NP 1741 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[to]{W}<2>\NP{V*}){W}){_}/NP{V}<3>){_} 1742 | 1 mod %f %l 1743 | 2 acomp %l %f 1744 | 3 pobj %l %f 1745 | 1746 | ((((S\NP)\(S\NP))/((S\NP)\(S\NP)))/(((S\NP)\(S\NP))/((S\NP)\(S\NP))))\(S[adj]\NP) 1747 | 2 (((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}/(((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}<2>){_}\(S[adj]{U}<1>\NP{T}){U}){_} 1748 | 1 mod %f %l 1749 | 2 mod %f %l 1750 | 1751 | ((S[b]\NP)/(S[pss]\NP))/NP 1752 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1753 | 1 nsubj %l %f 1754 | 2 acomp %l %f 1755 | 3 dobj %l %f 1756 | 1757 | ((S[adj]\NP)/(S[adj]\NP))/N 1758 | 2 (((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}<1>\NP{Z}){Y}){_}/N{W}<2>){_} 1759 | 1 mod %f %2 1760 | 2 det %f %l =fulldet 1761 | 2 mod %f %l 1762 | 1763 | N/S[for] 1764 | 1 (N{_}/S[for]{Y}<1>){_} 1765 | 1 ccomp %l %f 1766 | 1767 | (S\S)/PP 1768 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/PP{Z}<2>){_} 1769 | 1 mod %f %l 1770 | 2 dep %l %f =PP/NP 1771 | 1772 | (((S\NP)\(S\NP))/(S[pt]\NP))/NP 1773 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[pt]{W}<2>\NP{V}){W}){_}/NP{U}<3>){_} 1774 | 1 advcl %l %f 1775 | 2 ccomp %l %f 1776 | 3 ignore 1777 | # coindex the NPs to get the subject? 1778 | 1779 | ((S[dcl]\NP)/(S[adj]\NP))/PP 1780 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_}/PP{W}<3>){_} 1781 | 1 nsubj %l %f 1782 | 2 acomp %l %f 1783 | 3 prep %l %f =PP/NP 1784 | 1785 | ((S[b]\NP)/(S[to]\NP))/(S[adj]\NP) 1786 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/(S[adj]{V}<3>\NP{U}){V}){_} 1787 | 1 nsubj %l %f 1788 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 1789 | 3 acomp %l %f 1790 | 1791 | (S/S)/(S[asup]\NP) 1792 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/(S[asup]{Z}<2>\NP{W}){Z}){_} 1793 | 1 mod %f %l 1794 | 2 acomp %l %f 1795 | 1796 | # Inverted aux 1797 | (S[q]/(S[ng]\NP))/NP 1798 | 2 ((S[q]{_}/(S[ng]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 1799 | ! ((S[q]{Y}/(S[ng]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 1800 | 1 aux %f %l 1801 | 2 ignore 1802 | 1803 | ((S[pt]\NP)/(S[b]\NP))/NP 1804 | 3 (((S[pt]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1805 | 1 nsubj %l %f 1806 | 2 acomp %l %f 1807 | 3 ignore 1808 | 1809 | (S[dcl]\(S[pss]\NP))/NP 1810 | 2 ((S[dcl]{_}\(S[pss]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 1811 | 1 subj %l %f 1812 | 2 acomp %l %f =be 1813 | 2 dobj %l %f 1814 | 1815 | (((S[dcl]\NP)/(S[to]\NP))/(S[adj]\NP))/NP[expl] 1816 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/(S[adj]{V}<3>\NP{U}){V}){_}/NP[expl]{T}<4>){_} 1817 | 1 nsubj %l %f 1818 | 2 xcomp %3 %k =(S[to]\NP)/(S[b]\NP) # see "I find it hard to conceive" 1819 | 3 acomp %l %f 1820 | 4 dobj %l %f 1821 | 1822 | (PP/PP)/(PP/PP) 1823 | 1 ((PP{Y}/PP{Z}){W}/(PP{Y}/PP{Z}){W}<1>){_} 1824 | 1 mod %f %l 1825 | 1826 | ((NP\NP)\NP)/NP 1827 | 3 (((NP{Y}\NP{Y}<1>){_}\NP{Z}<2>){_}/NP{W}<3>){_} 1828 | 1 mod %f %l 1829 | 2 mod %l %f # "an hour outside Tokyo" 1830 | 3 pobj %l %f 1831 | 1832 | # Inverted aux 1833 | (S[q]/(S[adj]\NP))/NP 1834 | 2 ((S[q]{_}/(S[adj]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 1835 | 1 cop %f %l 1836 | 2 ignore 1837 | 1838 | ((S[pt]\NP)/(S[ng]\NP))/NP 1839 | 3 (((S[pt]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1840 | 1 nsubj %l %f 1841 | 2 acomp %l %f 1842 | 3 dobj %l %f 1843 | 1844 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/PP 1845 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_}/PP{V}<2>){_} 1846 | 1 mod %f %l 1847 | 2 dep %l %f =PP/NP # note that this category is sometimes used for "as soon as" which Stanford parser categorizes as advmod 1848 | 1849 | S\NP 1850 | 1 (S{_}\NP{Y}<1>){_} 1851 | 1 ignore # really odd analyses in CCGbank for this category 1852 | 1853 | ((S[b]\NP)/S[qem])/NP 1854 | 3 (((S[b]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_}/NP{W}<3>){_} 1855 | 1 nsubj %l %f 1856 | 2 ccomp %l %c =S[em]/S[dcl] 1857 | 3 dobj %l %f 1858 | 1859 | ((S[b]\NP)/S[em])/NP 1860 | 3 (((S[b]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/NP{W}<3>){_} 1861 | 1 nsubj %l %f 1862 | 2 ccomp %l %c =S[em]/S[dcl] 1863 | 3 dobj %l %f 1864 | 1865 | (S[adj]\NP)/(S[ng]\NP) 1866 | 2 ((S[adj]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 1867 | 1 ignore 1868 | 2 acomp %l %f 1869 | 1870 | (NP/(S[dcl]/NP))/N 1871 | 2 ((NP{Y}/(S[dcl]{Z}<1>/NP{Y*}){Z}){_}/N{Y}<2>){_} 1872 | 1 ignore # object gets picked up through co-indexing 1873 | 2 det %f %l 1874 | 1875 | ((NP\NP)\(NP\NP))/(S[dcl]\NP) 1876 | 2 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}<2>){_}/(S[dcl]{W}<1>\NP{V}){W}){_} 1877 | 1 acomp %l %f 1878 | 2 mod %f %l 1879 | 1880 | ((NP[nb]/N)\(NP[nb]/N))/(NP[nb]/N) 1881 | 3 (((NP[nb]{Y}/N{Y}<1>){_}\(NP[nb]{Z}/N{Z}){W}<2>){_}/(NP[nb]{V}/N{V}){U}<3>){_} 1882 | 1 det %f %l # not sure what to do with this one "up to 19 years" ; we probably code differently in the bio corpus anyway 1883 | 2 mod %f %l 1884 | 3 mod %f %l 1885 | 1886 | (N/N)\NP 1887 | 2 ((N{Y}/N{Y}<1>){_}\NP{Z}<2>){_} 1888 | 1 mod %f %l 1889 | 2 mod %f %l 1890 | 1891 | ((N/N)\(N/N))/(S[adj]\NP) 1892 | 2 (((N{Y}/N{Y}){Z}\(N{Y}/N{Y}){Z}<2>){_}/(S[adj]{W}<1>\NP{V}){W}){_} 1893 | 1 acomp %l %f 1894 | 2 mod %f %l 1895 | 1896 | (S[pt]\NP)/(S[b]\NP) 1897 | 2 ((S[pt]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 1898 | 1 nsubj %l %f 1899 | 2 acomp %l %f 1900 | 1901 | (S[pss]\NP)/S[dcl] 1902 | 2 ((S[pss]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_} 1903 | 1 nsubjpass %l %f 1904 | 2 ccomp %l %f 1905 | 1906 | ((S\NP)\(S\NP))/S[poss] 1907 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S[poss]{W}<2>){_} 1908 | 1 advcl %f %l 1909 | 2 ccomp %l %f 1910 | 1911 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/((S\NP)\(S\NP)) 1912 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){_}\((S{Y}\NP{Z}){Y}\(S{Y}\NP{Z}){Y}){W}<1>){_}/((S{U}\NP{Z}){U}\(S{U}\NP{Z}){U}){V}<2>){_} 1913 | 1 mod %f %l # not sure what to do with modifiers as complements, so currently just assigning ncmod (which is becoming a bucket category) 1914 | 2 mod %f %l 1915 | 1916 | ((S[dcl]\NP)/S[qem])/NP 1917 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_}/NP{W}<3>){_} 1918 | 1 nsubj %l %f 1919 | 2 ccomp %l %f 1920 | 3 dobj %l %f 1921 | 1922 | PP/S[dcl] 1923 | 1 (PP{_}/S[dcl]{Y}<1>){_} 1924 | 1 ccomp %l %f 1925 | 1926 | (S[pss]\NP)/S[qem] 1927 | 2 ((S[pss]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 1928 | 1 nsubjpass %l %f 1929 | 2 ccomp %l %f 1930 | 1931 | (((S\NP)\(S\NP))/(S[pss]\NP))/NP 1932 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[pss]{W}<2>\NP{V}){W}){_}/NP{U}<3>){_} 1933 | 1 advcl %l %f 1934 | 2 ccomp %l %f 1935 | 3 ignore 1936 | # coindex the NPs to get the subject? 1937 | 1938 | (S[ng]\NP)/S[for] 1939 | 2 ((S[ng]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_} 1940 | 1 nsubj %l %f 1941 | 2 ccomp %l %f 1942 | 1943 | (S[ng]\NP)/(S[b]\NP) 1944 | 2 ((S[ng]{_}\NP{Y}<1>){_}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 1945 | 1 nsubj %l %f 1946 | 2 acomp %l %f 1947 | 1948 | (S[dcl]\NP)/S[wq] 1949 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[wq]{Z}<2>){_} 1950 | 1 nsubj %l %f 1951 | 2 ccomp %l %f 1952 | 1953 | ((S[dcl]\NP)\NP)/(S[b]\NP) 1954 | 3 (((S[dcl]{_}\NP{Y}<1>){_}\NP{Z}<2>){_}/(S[b]{W}<3>\NP{V}){W}){_} 1955 | ! (((S[dcl]{W}\NP{Y}<1>){W}\NP{Z}<2>){W}/(S[b]{W}<3>\NP{V}){W}){_} 1956 | 1 ignore # ? . odd analysis in CCGbank; these are just auxiliaries 1957 | 2 ignore # ? 1958 | 3 aux %l %f 1959 | 1960 | (S[dcl]/NP)/NP 1961 | 2 ((S[dcl]{_}/NP{Y}<1>){_}/NP{Z}<2>){_} 1962 | 1 dobj %l %f 1963 | 2 nsubjpass %l %f 1964 | 1965 | ((NP\NP)\(NP\NP))/S[dcl] 1966 | 2 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}<2>){_}/S[dcl]{W}<1>){_} 1967 | 1 ccomp %l %f 1968 | 2 mod %f %l 1969 | 1970 | ((S/S)/(S[pt]\NP))/NP 1971 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/(S[pt]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1972 | 1 advcl %f %l 1973 | 2 ccomp %l %f 1974 | 3 ignore 1975 | 1976 | ((S[ng]\NP)/PP)/(S[adj]\NP) 1977 | 3 (((S[ng]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{Y*}){W}){_} 1978 | 1 nsubj %l %f 1979 | 2 prep %l %f =PP/NP 1980 | 3 acomp %l %f 1981 | 1982 | (S[dcl]\(S[b]\NP))/NP 1983 | 2 ((S[dcl]{_}\(S[b]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 1984 | 1 xsubj %l %f 1985 | 2 dobj %l %f 1986 | 1987 | ((S[dcl]\NP)/(S[to]\NP))/(S[adj]\NP) 1988 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/(S[adj]{V}<3>\NP{U}){V}){_} 1989 | 1 nsubj %l %f 1990 | 2 ignore # odd analyses in CCGbank 1991 | 3 acomp %l %f 1992 | 1993 | ((S[b]\NP)/(S[pt]\NP))/NP 1994 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 1995 | 1 nsubj %l %f 1996 | 2 acomp %l %f 1997 | 3 dobj %l %f 1998 | 1999 | ((S[b]\NP)/S[for])/(S[adj]\NP) 2000 | 3 (((S[b]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2001 | 1 nsubj %l %f 2002 | 2 ccomp %l %f 2003 | 3 acomp %l %f 2004 | 2005 | ((S/S)/(S[adj]\NP))/NP 2006 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_}/NP{V}<3>){_} 2007 | 1 mod %f %l 2008 | 2 acomp %l %f 2009 | 3 pobj %l %f 2010 | 2011 | (S[pss]\NP)/(S[ng]\NP) 2012 | 2 ((S[pss]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 2013 | 1 nsubjpass %l %f 2014 | 2 acomp %l %f 2015 | 2016 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/(S[adj]\NP) 2017 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<1>){_}/(S[adj]{V}<2>\NP{U}){V}){_} 2018 | 1 mod %f %l 2019 | 2 acomp %l %f 2020 | 2021 | ((S[dcl]\NP[expl])/S[em])/NP 2022 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/S[em]{Z}<2>){_}/NP{W}<3>){_} 2023 | 1 nsubj %l %f 2024 | 2 ccomp %l %c =S[em]/S[dcl] 2025 | 3 dobj %l %f 2026 | 2027 | ((S/S)/(S/S))/NP 2028 | 2 (((S[X]{Y}/S[X]{Y}){Z}/(S[X]{Y}/S[X]{Y}){Z}<1>){_}/NP{W}<2>){_} 2029 | 1 mod %f %l 2030 | 2 pobj %l %f 2031 | 2032 | (S\S)/S[q] 2033 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/S[q]{Z}<2>){_} 2034 | 1 advcl %f %l 2035 | 2 ccomp %l %f 2036 | 2037 | ((S/S)/(S[b]\NP))/NP 2038 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/(S[b]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2039 | 1 advcl %f %l 2040 | 2 ccomp %l %f 2041 | 3 ignore # co-indexing gets subj 2042 | 2043 | ((S/S)\NP)/NP 2044 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}\NP{Z}<2>){_}/NP{W}<3>){_} 2045 | 1 mod %f %l 2046 | 2 mod %f %l 2047 | 3 pobj %l %f 2048 | 2049 | (S[pt]\NP)/S[qem] 2050 | 2 ((S[pt]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 2051 | 1 nsubj %l %f 2052 | 2 ccomp %l %f 2053 | 2054 | (S[pss]\NP)/(S[pss]\NP) 2055 | 2 ((S[pss]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 2056 | 1 nsubjpass %l %f 2057 | 2 acomp %l %f 2058 | 2059 | (S[dcl]\NP[expl])/(S[pt]\NP) 2060 | 2 ((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[pt]{Z}<2>\NP{W}){Z}){_} 2061 | ! ((S[dcl]{Z}\NP[expl]{Y}<1>){Z}/(S[pt]{Z}<2>\NP{W}){Z}){_} 2062 | 1 ignore =aux 2063 | 1 nsubj %l %f 2064 | 2 aux %f %l =aux 2065 | 2 acomp %l %f 2066 | 2067 | ((S[dcl]\NP[expl])/S[em])/PP 2068 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/S[em]{Z}<2>){_}/PP{W}<3>){_} 2069 | 1 nsubj %l %f 2070 | 2 ccomp %l %c =S[em]/S[dcl] 2071 | 3 prep %l %f =PP/NP 2072 | 2073 | ((S[b]\NP)/S[dcl])/NP 2074 | 3 (((S[b]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_}/NP{W}<3>){_} 2075 | 1 nsubj %l %f 2076 | 2 ccomp %l %f 2077 | 3 dobj %l %f 2078 | 2079 | (S[adj]\NP)/S[for] 2080 | 2 ((S[adj]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_} 2081 | 1 ignore 2082 | 2 ccomp %l %f 2083 | 2084 | ((S[adj]\NP)\NP)/PP 2085 | 3 (((S[adj]{_}\NP{Y}<1>){_}\NP{Z}<2>){_}/PP{W}<3>){_} 2086 | 1 ignore 2087 | 2 mod %f %l 2088 | 3 prep %l %f =PP/NP 2089 | 2090 | (PP\PP)/N 2091 | 2 ((PP{Y}\PP{Y}<1>){_}/N{Z}<2>){_} 2092 | 1 mod %f %l 2093 | 2 det %f %l =fulldet 2094 | 2 mod %f %l 2095 | 2096 | ((NP\NP)\(NP\NP))\NP 2097 | 2 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}<2>){_}\NP{W}<1>){_} 2098 | 1 mod %f %l 2099 | 2 mod %f %l 2100 | 2101 | # wh-object determiner 2102 | (S[wq]/(S[q]/NP))/N 2103 | 2 ((S[wq]{_}/(S[q]{Y}<1>/NP{Z*}){Y}){_}/N{Z}<2>){_} 2104 | 1 ignore 2105 | 2 det %f %l 2106 | 2107 | ((S/S)\NP)/S[dcl] 2108 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}\NP{Z}<2>){_}/S[dcl]{W}<3>){_} 2109 | 1 mod %f %l 2110 | 2 mod %f %l 2111 | 3 ccomp %l %f 2112 | 2113 | ((S[qem]/(S[dcl]\NP))/N)/(S[adj]\NP) 2114 | 3 (((S[qem]{_}/(S[dcl]{Y}<1>\NP{Z}){Y}){_}/N{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2115 | 1 mod %f %l # consistent with ((S[qem]/(S[dcl]/NP))/N)/(S[adj]\NP) 2116 | 2 mod %3 %f 2117 | 3 dobj %l %f # wondering about det, but this is probably right 2118 | 2119 | ((S\NP)\(S\NP))/S 2120 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S{W}<2>){_} 2121 | 1 mod %f %l 2122 | 2 mod %f %l 2123 | 2124 | ((S[ng]\NP)/NP)/PP 2125 | 3 (((S[ng]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/PP{W}<3>){_} 2126 | 1 nsubj %l %f 2127 | 2 dobj %l %f 2128 | 3 prep %l %f =PP/NP 2129 | 2130 | S[dcl]/S[inv] 2131 | 1 (S[dcl]{_}/S[inv]{Y}<1>){_} 2132 | 1 nsubj %f %l # consistent with B+C's analysis 2133 | 2134 | (((S[b]\NP)/S[for])/(S[adj]\NP))/NP[expl] 2135 | 4 ((((S[b]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_}/NP[expl]{U}<4>){_} 2136 | 1 nsubj %l %f 2137 | 2 ccomp %3 %f # consistent with B+C 2138 | 3 acomp %l %f 2139 | 4 dobj %l %f 2140 | 2141 | (((S[adj]\NP)/(S[adj]\NP))/((S[adj]\NP)/(S[adj]\NP)))/(((S[adj]\NP)/(S[adj]\NP))/((S[adj]\NP)/(S[adj]\NP))) 2142 | 1 ((((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}/((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}){V}/(((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}/((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}){V}<1>){_} 2143 | 1 mod %f %l 2144 | 2145 | NP/((S[to]\NP)/NP) 2146 | 1 (NP{_}/((S[to]{Y}<1>\NP{Z}){Y}/NP{W}){Y}){_} 2147 | ! (NP{_}/((S[to]{Y}<1>\NP{Z}){Y}/NP{_}){Y}){_} 2148 | 1 ignore 2149 | # might want to co-index NP in the non-GR case 2150 | 2151 | (NP/(S[dcl]\NP))/N 2152 | 2 ((NP{_}/(S[dcl]{Y}<1>\NP{Z*}){Y}){_}/N{Z}<2>){_} 2153 | 1 ignore # co-indexing gets subj 2154 | 2 pobj %l %f # should be det here and in B+C markedup? 2155 | 2156 | (NP/NP)/(S[adj]\NP) 2157 | 2 ((NP{Y}/NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 2158 | 1 mod %f %2 # guess at what B+C would have 2159 | 2 mod %f %l 2160 | 2161 | ((NP[nb]/N)/PP)/(S[adj]\NP) 2162 | 3 (((NP[nb]{Y}/N{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2163 | 1 ignore # odd analysis in CCGbank 2164 | 2 mod %3 %f # consistent with B+C "as many as ..." 2165 | 3 pobj %l %f 2166 | 2167 | (N\N)/(N\N) 2168 | 1 ((N{Y}\N{Y}){Z}/(N{Y}\N{Y}){Z}<1>){_} 2169 | 1 mod %f %l 2170 | 2171 | (S\S)/(S[ng]\NP) 2172 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/(S[ng]{Z}<2>\NP{W}){Z}){_} 2173 | 1 prep %f %l 2174 | 2 pcomp %l %f 2175 | # "in explaining the name change" 2176 | # not sure this is always the correct analysis 2177 | 2178 | ((S/S)/PP)/NP 2179 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 2180 | 1 mod %f %l 2181 | 2 mod %3 %f # no egs; "with fast-food outlets on every corner" 2182 | 3 pobj %l %f 2183 | 2184 | ((S[qem]/(S[dcl]/NP))/N)/(S[adj]\NP) 2185 | 3 (((S[qem]{_}/(S[dcl]{Y}<1>/NP{Z*}){Y}){_}/N{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2186 | 1 mod %f %l 2187 | 2 mod %3 %f 2188 | 3 dobj %l %f # "how many more of these shocks the small investor can stand" 2189 | 2190 | ((S\NP)\(S\NP))/S[ng] 2191 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/S[ng]{W}<2>){_} 2192 | 1 advcl %f %l 2193 | 2 ccomp %l %f 2194 | 2195 | (((S\NP)\(S\NP))/NP)/NP 2196 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_}/NP{V}<3>){_} 2197 | 1 mod %f %l 2198 | 2 iobj %l %f # not sure what to do with this argument 2199 | 3 pobj %l %f 2200 | 2201 | ((S\NP)/(S\NP))\NP 2202 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_} 2203 | 1 mod %f %2 # guess at what B+C would have 2204 | 2 mod %f %l 2205 | 2206 | (S\NP)/NP 2207 | 2 ((S{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 2208 | 1 mod %f %l # most of these cases are prepostions in CCGbank 2209 | 2 dobj %l %f # not sure about this one, eg. "60 of them in retail", would it be dobj or pobj - not sure how to search for examples 2210 | 2211 | ((S[ng]\NP)/(S[pss]\NP))/NP 2212 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2213 | 1 nsubj %l %f 2214 | 2 acomp %l %f 2215 | 3 dobj %l %f 2216 | 2217 | S[frg]/NP 2218 | 1 (S[frg]{_}/NP{Y}<1>){_} 2219 | 1 dobj %l %f 2220 | 2221 | (S[dcl]\S[wq])/NP 2222 | 2 ((S[dcl]{_}\S[wq]{Y}<1>){_}/NP{Z}<2>){_} 2223 | 1 csubj %l %f 2224 | 2 dobj %l %f # this is like "xyz, asks Ms. Browning": shouldn't subj and obj be reversed? no exx in b&c 2225 | 2226 | ((S[b]\NP)/(S[to]\NP))/PP 2227 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{V}<3>){_} 2228 | 1 nsubj %l %f 2229 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2230 | 3 prep %l %f =PP/NP 2231 | 2232 | ((S[adj]\NP)\(S[adj]\NP))/(S[pss]\NP) 2233 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/(S[pss]{W}<2>\NP{V}){W}){_} 2234 | 1 mod %f %l 2235 | 2 acomp %l %f 2236 | 2237 | (NP\NP)/(S[to]\NP) 2238 | 2 ((NP{Y}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_} 2239 | 1 mod %f %l 2240 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2241 | 2242 | ((S/S)/S[dcl])/(S[adj]\NP) 2243 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/S[dcl]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2244 | 1 mod %f %l 2245 | 2 mod %f %3 # guess at what B+C would have 2246 | 3 dobj %l %f # probably OK - this is "the" and "sooner" in "the sooner it arrives" 2247 | 2248 | (S\S)/(S[dcl]\NP) 2249 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/(S[dcl]{Z}<2>\NP{W}){Z}){_} 2250 | 1 mod %f %l 2251 | 2 acomp %l %f 2252 | 2253 | ((S[pss]\NP)/PP)/(S[adj]\NP) 2254 | 3 (((S[pss]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2255 | 1 nsubjpass %l %f 2256 | 2 prep %l %f =PP/NP 2257 | 3 acomp %l %f 2258 | 2259 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/(S[pss]\NP) 2260 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<2>){_}/(S[pss]{V}<1>\NP{U}){V}){_} 2261 | 1 acomp %l %f 2262 | 2 mod %f %l 2263 | 2264 | ((((S\NP)\(S\NP))/((S\NP)\(S\NP)))/(((S\NP)\(S\NP))/((S\NP)\(S\NP))))/(S[asup]\NP) 2265 | 2 (((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}/(((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}){V}<2>){_}/(S[asup]{U}<1>\NP{T}){U}){_} 2266 | 1 acomp %l %f 2267 | 2 mod %f %l 2268 | 2269 | ((S\NP)/(S\NP))\((S\NP)/(S\NP)) 2270 | 1 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}\NP{Z}){Y}){W}<1>){_} 2271 | 1 mod %f %l 2272 | 2273 | ((S\NP)/(S\NP))/(S[asup]\NP) 2274 | 2 (((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[asup]{W}<2>\NP{V}){W}){_} 2275 | 1 mod %f %l 2276 | 2 acomp %l %f 2277 | 2278 | ((S[ng]\NP)/S[em])/NP 2279 | 3 (((S[ng]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/NP{W}<3>){_} 2280 | 1 nsubj %l %f 2281 | 2 ccomp %l %c =S[em]/S[dcl] 2282 | 3 dobj %l %f 2283 | 2284 | ((S[ng]\NP)/NP)/(S[adj]\NP) 2285 | 3 (((S[ng]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2286 | 1 nsubj %l %f 2287 | 2 dobj %l %f 2288 | 3 acomp %l %f 2289 | 2290 | (S[dcl]/(S[pt]\NP))/NP 2291 | 2 ((S[dcl]{_}/(S[pt]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 2292 | ! ((S[dcl]{Y}/(S[pt]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 2293 | 1 aux %f %l 2294 | 2 ignore # subj got by co-indexing 2295 | 2296 | (((S[dcl]\NP)/S[em])/(S[adj]\NP))/NP[expl] 2297 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_}/NP[expl]{U}<4>){_} 2298 | 1 nsubj %l %f 2299 | 2 ccomp %3 %c =S[em]/S[dcl] # "make it clear that" 2300 | 3 acomp %l %f 2301 | 4 dobj %l %f 2302 | 2303 | (S[dcl]\NP[expl])/(S[pss]\NP) 2304 | 2 ((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[pss]{Z}<2>\NP{W}){Z}){_} 2305 | ! ((S[dcl]{Z}\NP[expl]{Y}<1>){Z}/(S[pss]{Z}<2>\NP{Y}){Z}){_} 2306 | 1 ignore # nsubjpass from unification of NPs; Stanford doesn't mark "it" expletives specially 2307 | 2 auxpass %f %l 2308 | # "it is considered" 2309 | 2310 | (((S[b]\NP)/(S[to]\NP))/(S[adj]\NP))/NP[expl] 2311 | 4 ((((S[b]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/(S[adj]{V}<3>\NP{U}){V}){_}/NP[expl]{T}<4>){_} 2312 | 1 nsubj %l %f 2313 | 2 xcomp %3 %k =(S[to]\NP)/(S[b]\NP) # "make it hard to find" 2314 | 3 acomp %l %f 2315 | 4 dobj %l %f 2316 | 2317 | S[adj]/S[adj] 2318 | 1 (S[adj]{Y}/S[adj]{Y}<1>){_} 2319 | 1 mod %f %l 2320 | 2321 | (PP\PP)/S[dcl] 2322 | 2 ((PP{Y}\PP{Y}<1>){_}/S[dcl]{Z}<2>){_} 2323 | 1 mod %f %l 2324 | 2 ccomp %l %f 2325 | 2326 | (NP\NP)/S[inv] 2327 | 2 ((NP{Y}\NP{Y}<1>){_}/S[inv]{Z}<2>){_} 2328 | 1 mod %f %l 2329 | 2 ccomp %l %f 2330 | 2331 | (NP\NP)/S 2332 | 2 ((NP{Y}\NP{Y}<1>){_}/S{Z}<2>){_} 2333 | 1 mod %f %l 2334 | 2 mod %f %l # not sure about this one 2335 | 2336 | N/NP 2337 | 1 (N{Y}/NP{Y}<1>){_} 2338 | 1 dobj %l %f 2339 | 2340 | ((S/S)/(S[to]\NP))/NP 2341 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/(S[to]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2342 | 1 mod %f %l 2343 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) # "for its employees to sign ..." 2344 | 3 ignore 2345 | 2346 | (S\S)/(S[b]\NP) 2347 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/(S[b]{Z}<2>\NP{W}){Z}){_} 2348 | 1 mod %f %l 2349 | 2 acomp %l %f 2350 | 2351 | (S\S)/N 2352 | 2 ((S[X]{Y}\S[X]{Y}<1>){_}/N{Z}<2>){_} 2353 | 1 mod %f %2 2354 | 2 det %f %l =fulldet 2355 | 2 mod %f %l 2356 | 2357 | ((S[pt]\NP)/S[em])/NP 2358 | 3 (((S[pt]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/NP{W}<3>){_} 2359 | 1 nsubj %l %f 2360 | 2 ccomp %l %c =S[em]/S[dcl] 2361 | 3 dobj %l %f 2362 | 2363 | (((S\NP)/(S\NP))\((S\NP)/(S\NP)))/NP 2364 | 2 ((((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}\NP{Z}){Y}){W}<2>){_}/NP{V}<1>){_} 2365 | 1 pobj %l %f 2366 | 2 mod %f %l 2367 | 2368 | ((S[ng]\NP)/(S[to]\NP))/PP 2369 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{W}<3>){_} 2370 | 1 nsubj %l %f 2371 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2372 | 3 prep %l %f =PP/NP 2373 | 2374 | (S[dcl]\(S[b]\NP))\NP 2375 | 2 ((S[dcl]{_}\(S[b]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_} 2376 | 1 mod %f %l # not sure about this " ... , Mr Weiss added" 2377 | 2 nsubj %l %f 2378 | 2379 | (S[dcl]\S[b])\NP 2380 | 2 ((S[dcl]{_}\S[b]{Y}<1>){_}\NP{Z}<2>){_} 2381 | 1 mod %f %l # not sure about this " ... , Mr Bickwit said" 2382 | 2 nsubj %l %f 2383 | 2384 | ((S[dcl]\NP)/(S[pt]\NP))/NP 2385 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2386 | 1 nsubj %l %f 2387 | 2 acomp %l %f 2388 | 3 dobj %l %f 2389 | 2390 | ((S[dcl]\NP)/S[em])/PP 2391 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/PP{W}<3>){_} 2392 | 1 nsubj %l %f 2393 | 2 ccomp %l %c =S[em]/S[dcl] 2394 | 3 prep %l %f =PP/NP 2395 | 2396 | ((S[dcl]\NP[expl])/(S[to]\NP))/PP 2397 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{V}<3>){_} 2398 | 1 nsubj %l %f 2399 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2400 | 3 prep %l %f =PP/NP 2401 | 2402 | ((S[adj]\NP)/S[dcl])/(S[adj]\NP) 2403 | 3 (((S[adj]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2404 | 1 ignore 2405 | 2 mod %f %l # not sure about this 2406 | 3 mod %f %l 2407 | 2408 | (NP\NP)/S[q] 2409 | 2 ((NP{Y}\NP{Y}<1>){_}/S[q]{Z}<2>){_} 2410 | 1 advcl %f %2 2411 | 2 mod %f %l # consistent with (NP\NP)/S[qem] 2412 | 2413 | ((NP\NP)\(NP\NP))/(NP\NP) 2414 | 2 (((NP{Y}\NP{Y}){_}\(NP{Z}\NP{Z}){W}<1>){_}/(NP{V}\NP{V}){U}<2>){_} 2415 | 1 mod %f %l 2416 | 2 mod %f %l # another "% to %" type case; go for default option 2417 | 2418 | S[wq]/(S[b]\NP) 2419 | 1 (S[wq]{_}/(S[b]{Y}<1>\NP{Z}){Y}){_} 2420 | 1 acomp %l %f 2421 | 2422 | ((S/S)/(S[pss]\NP))/NP 2423 | 3 (((S[X]{Y}/S[X]{Y}<1>){_}/(S[pss]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2424 | 1 advcl %f %l 2425 | 2 ccomp %l %f 2426 | 3 ignore # co-indexing gets subj 2427 | 2428 | S/(S[dcl]\NP) 2429 | 1 (S{_}/(S[dcl]{Y}<1>\NP{Z}){Y}){_} 2430 | ! (S{_}/(S[dcl]{Y}<1>\NP{_}){Y}){_} 2431 | 1 ignore 2432 | 2433 | (((S\NP)\(S\NP))/(S[adj]\NP))/NP 2434 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[adj]{W}<2>\NP{V*}){W}){_}/NP{V}<3>){_} 2435 | 1 mod %f %l 2436 | 2 acomp %l %f 2437 | 3 pobj %l %f # not perfect; few exx in ccgbank and one is a verb 2438 | 2439 | (((S\NP)\(S\NP))/PP)/(S[adj]\NP) 2440 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/PP{W}<2>){_}/(S[adj]{V}<3>\NP{U}){V}){_} 2441 | 1 mod %f %l 2442 | 2 mod %3 %f # "as much as ..." 2443 | 3 pobj %l %f 2444 | 2445 | (((S[ng]\NP)/PP)/PP)/NP 2446 | 4 ((((S[ng]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_}/NP{V}<4>){_} 2447 | 1 nsubj %l %f 2448 | 2 mod %f %l # don't have two iobj's 2449 | 3 iobj %l %f =PP/NP 2450 | 4 dobj %l %f 2451 | 2452 | (S[dcl]\S[wq])\NP 2453 | 2 ((S[dcl]{_}\S[wq]{Y}<1>){_}\NP{Z}<2>){_} 2454 | 1 mod %f %l # not sure about this " ... , he asked" 2455 | 2 nsubj %l %f 2456 | 2457 | (S[dcl]\NP)/S[q] 2458 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[q]{Z}<2>){_} 2459 | 1 nsubj %l %f 2460 | 2 ccomp %l %f 2461 | 2462 | ((S[dcl]\NP[expl])/S[dcl])/(S[adj]\NP) 2463 | 3 (((S[dcl]{_}\NP[expl]{Y}<1>){_}/S[dcl]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2464 | 1 nsubj %l %f 2465 | 2 ccomp %3 %f # "it is doubtful any collaboration will take place" 2466 | 3 acomp %l %f 2467 | 2468 | ((S[b]\NP)/S[for])/NP 2469 | 3 (((S[b]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_}/NP{W}<3>){_} 2470 | 1 nsubj %l %f 2471 | 2 ccomp %l %f 2472 | 3 acomp %l %f =be 2473 | 3 dobj %l %f 2474 | 2475 | ((S[adj]\NP)/(S[to]\NP))/PP 2476 | 3 (((S[adj]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_}/PP{W}<3>){_} 2477 | 1 ignore 2478 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2479 | 3 iobj %l %f =PP/NP 2480 | 2481 | (S[adj]\NP)/S[qem] 2482 | 2 ((S[adj]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_} 2483 | 1 ignore 2484 | 2 ccomp %l %f 2485 | 2486 | (((S[adj]\NP)\(S[adj]\NP))/((S[adj]\NP)\(S[adj]\NP)))/NP 2487 | 2 ((((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}\NP{Z}){Y}){W}/((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}\NP{Z}){Y}){W}<2>){_}/NP{V}<1>){_} 2488 | 1 pobj %l %f 2489 | 2 mod %f %l 2490 | 2491 | N/S[qem] 2492 | 1 (N{_}/S[qem]{Y}<1>){_} 2493 | 1 ccomp %l %f 2494 | 2495 | (NP\NP)/S[wq] 2496 | 2 ((NP{Y}\NP{Y}<1>){_}/S[wq]{Z}<2>){_} 2497 | 1 mod %f %l 2498 | 2 ccomp %l %f 2499 | 2500 | ((NP[nb]/N)/(NP[nb]/N))/(S[asup]\NP) 2501 | 2 (((NP[nb]{Y}/N{Y}){Z}/(NP[nb]{Y}/N{Y}){Z}<2>){_}/(S[asup]{W}<1>\NP{V}){W}){_} 2502 | 1 acomp %l %f 2503 | 2 mod %f %l 2504 | 2505 | ((N/N)/(N/N))/N 2506 | 2 (((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}<2>){_}/N{W}<1>){_} 2507 | 1 det %f %l =fulldet 2508 | 1 mod %f %l 2509 | 2 mod %f %1 2510 | 2511 | (S/S)/S[em] 2512 | 2 ((S{Y}/S{Y}<1>){_}/S[em]{Z}<2>){_} 2513 | 1 mod %f %l 2514 | 2 ccomp %l %c =S[em]/S[dcl] 2515 | 2516 | ((S[pt]\NP)/NP)/(S[adj]\NP) 2517 | 3 (((S[pt]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2518 | 1 nsubj %l %f 2519 | 2 dobj %l %f 2520 | 3 acomp %l %f 2521 | 2522 | ((S[ng]\NP)/S[dcl])/NP 2523 | 3 (((S[ng]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_}/NP{W}<3>){_} 2524 | 1 nsubj %l %f 2525 | 2 ccomp %l %f 2526 | 3 dobj %l %f 2527 | 2528 | S/N 2529 | 1 (S{_}/N{Y}<1>){_} 2530 | 1 det %f %l # odd analsyses in CCGbank 2531 | 2532 | ((S[dcl]\S[dcl])\NP)/(S[pt]\NP) 2533 | 3 (((S[dcl]{_}\S[dcl]{Y}<1>){_}\NP{Z}<2>){_}/(S[pt]{W}<3>\NP{Z*}){W}){_} 2534 | ! (((S[dcl]{W}\S[dcl]{Y}<1>){W}\NP{Z}<2>){W}/(S[pt]{W}<3>\NP{Z*}){W}){_} 2535 | 1 advcl %f %l 2536 | 2 nsubj %l %f 2537 | 3 aux %f %l 2538 | 2539 | ((S[dcl]\NP)\NP)/NP 2540 | 3 (((S[dcl]{_}\NP{Y}<1>){_}\NP{Z}<2>){_}/NP{W}<3>){_} 2541 | 1 ignore # odd analsyses in CCGbank 2542 | 2 nsubj %l %f 2543 | 3 dobj %l %f 2544 | 2545 | (S[b]\NP)/S[for] 2546 | 2 ((S[b]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_} 2547 | 1 nsubj %l %f 2548 | 2 ccomp %l %f 2549 | 2550 | (((S[b]\NP)/PP)/NP)/NP 2551 | 4 ((((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_}/NP{V}<4>){_} 2552 | 1 nsubj %l %f 2553 | 2 iobj %l %f =PP/NP 2554 | 3 iobj %l %f 2555 | 4 dobj %l %f # "give investors something for their money" 2556 | 2557 | ((S[adj]\NP)/(S[adj]\NP))/(S[asup]\NP) 2558 | 2 (((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}<1>\NP{Z}){Y}){_}/(S[asup]{W}<2>\NP{V}){W}){_} 2559 | 1 mod %f %2 # "were at least affordable" 2560 | 2 acomp %l %f 2561 | 2562 | N/PP 2563 | 1 (N{_}/PP{Y}<1>){_} 2564 | 1 iobj %l %f =PP/NP 2565 | 2566 | ((NP\NP)/((S[to]\NP)/NP))/NP 2567 | 4 (((NP{Y}\NP{Y}<1>){_}/((S[to]{Z}<2>\NP{W*}){Z}/NP{Y*}<3>){Z}){_}/NP{W}<4>){_} 2568 | 1 mod %f %l 2569 | 2 ccomp %l %f 2570 | 3 ignore 2571 | # one too many arguments here 2572 | 2573 | ((NP\NP)\NP)/PP 2574 | 3 (((NP{Y}\NP{Y}<1>){_}\NP{Z}<2>){_}/PP{W}<3>){_} 2575 | 1 mod %f %l 2576 | 2 mod %f %l 2577 | 3 iobj %l %f =PP/NP 2578 | 2579 | ((NP\NP)\(NP\NP))/N 2580 | 2 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}<2>){_}/N{W}<1>){_} 2581 | 1 pobj %l %f 2582 | 2 mod %f %1 2583 | 2584 | (NP/N)/(NP/N) 2585 | 1 ((NP{Y}/N{Y}){Z}/(NP{Y}/N{Y}){Z}<1>){_} 2586 | 1 mod %f %l 2587 | 2588 | (((N/N)/(N/N))\((N/N)/(N/N)))/((N/N)/(N/N)) 2589 | 2 ((((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){_}\((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){W}<1>){_}/((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){V}<2>){_} 2590 | 1 dep %f %l # following BioInfer 2591 | 2 mod %1 %f # e.g. "5 to 10", SD has a number dependency 2592 | 2593 | S[wq]/S[poss] 2594 | 1 (S[wq]{_}/S[poss]{Y}<1>){_} 2595 | 1 ccomp %f %l 2596 | 2597 | ((S/S)/(S/S))/((S/S)/(S/S)) 2598 | 1 (((S[X]{Y}/S[X]{Y}){Z}/(S[X]{Y}/S[X]{Y}){Z}){W}/((S[X]{Y}/S[X]{Y}){Z}/(S[X]{Y}/S[X]{Y}){Z}){W}<1>){_} 2599 | 1 mod %f %l 2600 | 2601 | ((S\S)\(S\S))/NP 2602 | 2 (((S[X]{Y}\S[X]{Y}){Z}\(S[X]{Y}\S[X]{Y}){Z}<2>){_}/NP{W}<1>){_} 2603 | 1 pobj %l %f 2604 | 2 mod %f %l 2605 | 2606 | (S/S)/(S[b]\NP) 2607 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/(S[b]{Z}<2>\NP{W}){Z}){_} 2608 | 1 mod %f %l 2609 | 2 acomp %l %f 2610 | 2611 | # Inverted aux 2612 | (S[q]/(S[pt]\NP))/NP 2613 | 2 ((S[q]{_}/(S[pt]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 2614 | ! ((S[q]{Y}/(S[pt]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 2615 | 1 aux %f %l 2616 | 2 ignore 2617 | 2618 | (S[q]/(S[dcl]\NP))/NP 2619 | 2 ((S[q]{_}/(S[dcl]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 2620 | ! ((S[q]{Y}/(S[dcl]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 2621 | 1 aux %f %l 2622 | 2 ignore 2623 | 2624 | S[q]/NP 2625 | 1 (S[q]{_}/NP{Y}<1>){_} 2626 | 1 nsubjpass %l %f 2627 | 2628 | ((S[qem]/S[dcl])\((NP\NP)/NP))/N 2629 | 3 (((S[qem]{_}/S[dcl]{Y}<1>){_}\((NP{Z}\NP{Z}){W}/NP{V}){U}<3>){_}/N{T}<2>){_} 2630 | 1 ccomp %l %f 2631 | 2 dobj %l %f # I think this should be det here and in B+C markedup 2632 | 3 mod %l %f # "detailing on what points member states ..." 2633 | 2634 | ((S[pt]\NP)/(S[to]\NP))/(S[adj]\NP) 2635 | 3 (((S[pt]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/(S[adj]{V}<3>\NP{U}){V}){_} 2636 | 1 nsubj %l %f 2637 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2638 | 3 acomp %l %f 2639 | 2640 | ((S[pt]\NP)/(S[pss]\NP))/NP 2641 | 3 (((S[pt]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2642 | 1 nsubj %l %f 2643 | 2 acomp %l %f 2644 | 3 dobj %l %f 2645 | 2646 | (((S\NP)\(S\NP))\((S\NP)\(S\NP)))/(S[dcl]\NP) 2647 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}\((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<2>){_}/(S[dcl]{V}<1>\NP{U}){V}){_} 2648 | 1 acomp %l %f 2649 | 2 mod %f %l 2650 | 2651 | (((S\NP)\(S\NP))/((S\NP)\(S\NP)))/N 2652 | 2 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}/((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}\NP{Z}){Y}){W}<2>){_}/N{V}<1>){_} 2653 | 1 pobj %l %f 2654 | 2 mod %f %1 # "a lot less" 2655 | 2656 | ((S\NP)\(S\NP))/(S[asup]\NP) 2657 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(S[asup]{W}<2>\NP{V}){W}){_} 2658 | 1 mod %f %l 2659 | 2 acomp %l %f 2660 | 2661 | (((S\NP)\(S\NP))\NP)/PP 2662 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_}/PP{V}<3>){_} 2663 | 1 mod %f %l 2664 | 2 mod %l %f # "30 miles Northeast of New York" 2665 | 3 iobj %l %f =PP/NP 2666 | 2667 | ((S[ng]\NP)/S[qem])/NP 2668 | 3 (((S[ng]{_}\NP{Y}<1>){_}/S[qem]{Z}<2>){_}/NP{W}<3>){_} 2669 | 1 nsubj %l %f 2670 | 2 ccomp %l %f 2671 | 3 dobj %l %f 2672 | 2673 | ((S[ng]\NP)/(S[pt]\NP))/NP 2674 | 3 (((S[ng]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{W*}){Z}){_}/NP{W}<3>){_} 2675 | 1 nsubj %l %f 2676 | 2 acomp %l %f 2677 | 3 dobj %l %f 2678 | 2679 | (((S[ng]\NP)/S[for])/(S[adj]\NP))/NP[expl] 2680 | 4 ((((S[ng]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_}/NP[expl]{U}<4>){_} 2681 | 1 nsubj %l %f 2682 | 2 ccomp %3 %f # consistent with (((S[b]\NP)/S[for])/(S[adj]\NP))/NP[expl] 2683 | 3 acomp %l %f 2684 | 4 dobj %l %f 2685 | 2686 | S[frg]/(S[adj]\NP) 2687 | 1 (S[frg]{_}/(S[adj]{Y}<1>\NP{Z}){Y}){_} 2688 | 1 acomp %l %f 2689 | 2690 | S[em]/(S[dcl]\NP) 2691 | 1 (S[em]{_}/(S[dcl]{Y}<1>\NP{Z}){Y}){_} 2692 | 1 acomp %l %f 2693 | 2694 | (S[dcl]\S[qem])/PP 2695 | 2 ((S[dcl]{_}\S[qem]{Y}<1>){_}/PP{Z}<2>){_} 2696 | 1 csubj %l %f 2697 | 2 iobj %l %f =PP/NP 2698 | 2 acomp %l %f =PP/(S[ng]\NP) 2699 | 2 acomp %l %f =PP/(S[adj]\NP) 2700 | 2701 | (S[dcl]\S[qem])/NP 2702 | 2 ((S[dcl]{_}\S[qem]{Y}<1>){_}/NP{Z}<2>){_} 2703 | 1 csubj %l %f # not sure about this "how many times we're going to feel good is another question" 2704 | 2 dobj %l %f 2705 | 2706 | ((S[dcl]\S[dcl])\NP)/PP 2707 | 3 (((S[dcl]{_}\S[dcl]{Y}<1>){_}\NP{Z}<2>){_}/PP{W}<3>){_} 2708 | 1 advcl %l %f 2709 | 2 nsubj %l %f 2710 | 3 iobj %l %f =PP/NP 2711 | 2712 | (S[dcl]\(S[dcl]\NP))/NP 2713 | 2 ((S[dcl]{_}\(S[dcl]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 2714 | 1 xsubj %l %f 2715 | 2 dobj %l %f 2716 | 2717 | (((S[dcl]\NP)/S[for])/(S[adj]\NP))/NP[expl] 2718 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/S[for]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_}/NP[expl]{U}<4>){_} 2719 | 1 nsubj %l %f 2720 | 2 ccomp %3 %f # consistent with (((S[b]\NP)/S[for])/(S[adj]\NP))/NP[expl] 2721 | 3 acomp %l %f 2722 | 4 dobj %l %f 2723 | 2724 | ((S[dcl]\NP)/S[em])/(S[adj]\NP) 2725 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_} 2726 | 1 nsubj %l %f 2727 | 2 ccomp %3 %c =S[em]/S[dcl] # "make clear that ..." 2728 | 3 acomp %l %f 2729 | 2730 | ((S[dcl]\NP)/S[dcl])/PP 2731 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/S[dcl]{Z}<2>){_}/PP{W}<3>){_} 2732 | 1 nsubj %l %f 2733 | 2 ccomp %l %f 2734 | 3 iobj %l %f =PP/NP 2735 | 2736 | (S[dcl]\NP)/S[b] 2737 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[b]{Z}<2>){_} 2738 | 1 nsubj %l %f 2739 | 2 ccomp %l %f 2740 | 2741 | (S[dcl]\NP)\NP 2742 | 2 ((S[dcl]{_}\NP{Y}<1>){_}\NP{Z}<2>){_} 2743 | 1 mod %f %l 2744 | 2 nsubj %l %f 2745 | 2746 | (S[dcl]\NP[expl])/(S[to]\NP) 2747 | 2 ((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_} 2748 | 1 nsubj %l %f 2749 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 2750 | 2751 | (S[b]\NP)/S[wq] 2752 | 2 ((S[b]{_}\NP{Y}<1>){_}/S[wq]{Z}<2>){_} 2753 | 1 nsubj %l %f 2754 | 2 ccomp %l %f 2755 | 2756 | ((S[b]\NP)/S[em])/PP 2757 | 3 (((S[b]{_}\NP{Y}<1>){_}/S[em]{Z}<2>){_}/PP{W}<3>){_} 2758 | 1 nsubj %l %f 2759 | 2 ccomp %l %c =S[em]/S[dcl] 2760 | 3 iobj %l %f =PP/NP 2761 | 2762 | (((S[b]\NP)/PP)/(S[adj]\NP))/NP 2763 | 4 ((((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_}/NP{U}<4>){_} 2764 | 1 nsubj %l %f 2765 | 2 iobj %l %f =PP/NP 2766 | 3 acomp %l %f 2767 | 4 dobj %l %f 2768 | 2769 | ((S[b]\NP)\NP)/(S[b]\NP) 2770 | 3 (((S[b]{_}\NP{Y}<1>){_}\NP{Z}<2>){_}/(S[b]{V}<3>\NP{U}){V}){_} 2771 | ! (((S[b]{V}\NP{Y}<1>){V}\NP{Z}<2>){V}/(S[b]{V}<3>\NP{U}){V}){_} 2772 | 1 ignore 2773 | 2 ignore # odd analysis 2774 | 3 aux %f %l 2775 | 2776 | ((S[b]\NP)\NP)/NP 2777 | 3 (((S[b]{_}\NP{Y}<1>){_}\NP{Z}<2>){_}/NP{W}<3>){_} 2778 | 1 ignore # odd analysis 2779 | 2 nsubj %l %f 2780 | 3 dobj %l %f 2781 | 2782 | S[as]/S[poss] 2783 | 1 (S[as]{_}/S[poss]{Y}<1>){_} 2784 | 1 ccomp %l %f 2785 | 2786 | ((S[adj]\NP)/((S[to]\NP)/NP))/(S[adj]\NP) 2787 | 2 (((S[adj]{Y}\NP{Z}){_}/((S[to]{W}<1>\NP{V}){W}/NP{U}){W}){_}/(S[adj]{Y}<2>\NP{Z}){Y}){_} 2788 | 1 xcomp %2 %k =(S[to]\NP)/(S[b]\NP) # "too small to maintain" 2789 | 2 mod %f %l # "too small" as ncmod rather than xcomp 2790 | 2791 | ((S[adj]\NP)\(S[adj]\NP))/S[inv] 2792 | 2 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}<1>\NP{Z}){Y}){_}/S[inv]{W}<2>){_} 2793 | 1 mod %f %l 2794 | 2 ccomp %l %f 2795 | 2796 | (NP\NP)/S[b] 2797 | 2 ((NP{Y}\NP{Y}<1>){_}/S[b]{Z}<2>){_} 2798 | 1 mod %f %l 2799 | 2 ccomp %l %f 2800 | 2801 | ((((NP\NP)\(NP\NP))/((NP\NP)\(NP\NP)))\(((NP\NP)\(NP\NP))/((NP\NP)\(NP\NP))))/(((NP\NP)\(NP\NP))/((NP\NP)\(NP\NP))) 2802 | 2 (((((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}/((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}){_}\(((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}/((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}){V}<1>){_}/(((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}/((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}){U}<2>){_} 2803 | 1 mod %f %l 2804 | 2 mod %f %l # default option 2805 | 2806 | ((NP\NP)\(NP\NP))\((NP\NP)\(NP\NP)) 2807 | 1 (((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}\((NP{Y}\NP{Y}){Z}\(NP{Y}\NP{Y}){Z}){W}<1>){_} 2808 | 1 mod %f %l 2809 | 2810 | ((NP\NP)/N)\NP 2811 | 3 (((NP{Y}\NP{Y}<1>){_}/N{Z}<2>){_}\NP{W}<3>){_} 2812 | 1 mod %f %3 2813 | 2 poss %f %3 2814 | 2815 | ((N\N)\(N\N))/NP 2816 | 2 (((N{Y}\N{Y}){Z}\(N{Y}\N{Y}){Z}<2>){_}/NP{W}<1>){_} 2817 | 1 pobj %l %f 2818 | 2 mod %f %l 2819 | 2820 | N/N[num] 2821 | 1 (N{_}/N[num]{Y}<1>){_} 2822 | 1 num %f %l # $ is head 2823 | 2824 | (N/N)/N[num] 2825 | 2 ((N{Y}/N{Y}<1>){_}/N[num]{Z}<2>){_} 2826 | 1 mod %f %l 2827 | 2 num %f %l 2828 | 2829 | S[bem]/S[b] 2830 | 1 (S[bem]{_}/S[b]{Y}<1>){_} 2831 | 1 ignore 2832 | 2833 | (S[dcl]\NP[thr])/(S[b]\NP) 2834 | 2 ((S[dcl]{_}\NP[thr]{Y}<1>){_}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 2835 | ! ((S[dcl]{Z}\NP[thr]{Y}<1>){Z}/(S[b]{Z}<2>\NP{Y*}){Z}){_} 2836 | 1 ignore =aux 2837 | 1 nsubj %l %f 2838 | 2 aux %f %l =aux 2839 | 2 acomp %l %f 2840 | 2841 | (S[dcl]\NP[thr])/(S[pt]\NP) 2842 | 2 ((S[dcl]{_}\NP[thr]{Y}<1>){_}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 2843 | ! ((S[dcl]{Z}\NP[thr]{Y}<1>){Z}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 2844 | 1 ignore =aux 2845 | 1 nsubj %l %f 2846 | 2 aux %f %l =aux 2847 | 2 acomp %l %f 2848 | 2849 | (S[dcl]\NP)/S[bem] 2850 | 2 ((S[dcl]{_}\NP{Y}<1>){_}/S[bem]{Z}<2>){_} 2851 | 1 nsubj %l %f 2852 | 2 ccomp %l %c =S[bem]/S[b] 2853 | 2854 | NP/(NP\NP) 2855 | 1 (NP{_}/(NP{Y}\NP{Y}){Z}<1>){_} 2856 | 1 ignore 2857 | # is this in 2-21? 2858 | 2859 | ((S\NP)\(S\NP))/N[num] 2860 | 2 (((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/N[num]{W}<2>){_} 2861 | 1 mod %f %l 2862 | 2 num %f %l # "Dec. 1" Dec. is the head 2863 | 2864 | (S[ng]\NP)/S[bem] 2865 | 2 ((S[ng]{_}\NP{Y}<1>){_}/S[bem]{Z}<2>){_} 2866 | 1 nsubj %l %f 2867 | 2 ccomp %l %c =S[bem]/S[b] 2868 | 2869 | (S/S)/S[inv] 2870 | 2 ((S[X]{Y}/S[X]{Y}<1>){_}/S[inv]{Z}<2>){_} 2871 | 1 advcl %f %l 2872 | 2 ccomp %l %f 2873 | 2874 | (S[dcl]\NP[thr])/NP 2875 | 2 ((S[dcl]{_}\NP[thr]{Y}<1>){_}/NP{Z}<2>){_} 2876 | 1 expl %l %f 2877 | 2 nsubj %l %f 2878 | # "there was little change", "there arose a great disturbance" 2879 | 2880 | (S[adj]\NP)/N[num] 2881 | 2 ((S[adj]{_}\NP{Y}<1>){_}/N[num]{Z}<2>){_} 2882 | 1 ignore 2883 | 2 num %f %l 2884 | 2885 | (NP\NP)/N[num] 2886 | 2 ((NP{Y}\NP{Y}<1>){_}/N[num]{Z}<2>){_} 2887 | 1 mod %f %l 2888 | 2 num %f %l 2889 | 2890 | # extra categories for questions 2891 | # Possessive in object question: what fruit's stone does ... come from 2892 | ((S[wq]/(S[q]/NP))/N)\(S[wq]/(S[q]/NP)) 2893 | 3 (((S[wq]{Y}/(S[q]{Z}<1>/NP{W}){Z}){Y}/N{V}<2>){Y}\(S[wq]{Y}<3>/(S[q]{Z*}/NP{W}){Z}){Y}){_} 2894 | 1 ignore 2895 | 2 poss %f %3 # needs to be post-processed in python 2896 | 3 possessive %f %l # needs to be post-processed in python 2897 | 2898 | # Possessive in subject question: what country's leader was ... 2899 | ((S[wq]/(S[dcl]\NP))/N)\(S[wq]/(S[dcl]\NP)) 2900 | 3 (((S[wq]{Y}/(S[dcl]{Z}<1>\NP{W}){Z}){Y}/N{V}<2>){Y}\(S[wq]{Y}<3>/(S[dcl]{Z*}\NP{W}){Z}){Y}){_} 2901 | ! (((S[wq]{Y}/(S[dcl]{Z}<1>\NP{V}){Z}){Y}/N{V}<2>){Y}\(S[wq]{Y}<3>/(S[dcl]{Z*}\NP{W}){Z}){Y}){_} 2902 | 1 ignore 2903 | 2 poss %f %3 # needs to be post-processed in python 2904 | 3 possessive %f %l # needs to be post-processed in python 2905 | # Or "what country" could have "country" as the head? What would this break? 2906 | 2907 | (S[wq]/(S[q]/NP))/NP 2908 | 2 ((S[wq]{_}/(S[q]{Y}<1>/NP{Z*}){Y}){_}/NP{Z}<2>){_} 2909 | 1 mod %f %l 2910 | 2 dobj %l %f # no exx in ccgbank 2911 | 2912 | ((S[wq]/(S[dcl]\NP))/N)/(NP/N) 2913 | 3 (((S[wq]{_}/(S[dcl]{Y}<1>\NP{Z*}){Y}){_}/N{Z}<2>){_}/(NP{W}/N{W}){V}<3>){_} 2914 | 1 mod %f %l 2915 | 2 dobj %l %f # no exx in ccgbank 2916 | 3 mod %2 %f 2917 | # based on "How many abbots can the abbey afford ?" from the RASP grammar development files 2918 | 2919 | ((S[wq]/S[q])\(S[wq]/S[q]))/NP 2920 | 2 (((S[wq]{Y}/S[q]{Y}){Z}\(S[wq]{Y}/S[q]{Y}){Z}<1>){_}/NP{W}<2>){_} 2921 | 1 mod %f %l 2922 | 2 pobj %l %f # no exx in ccgbank, so this is a guess 2923 | 2924 | # Adverbial PP 2925 | (S[wq]\S[wq])/NP 2926 | 2 ((S[wq]{Y}\S[wq]{Y}<1>){_}/NP{Z}<2>){_} 2927 | 1 prep %f %l 2928 | 2 pobj %l %f 2929 | 2930 | # VP modifier for question VP 2931 | ((S/NP)\(S/NP))/NP 2932 | 2 (((S[X]{Y}/NP{Z}){Y}\(S[X]{Y}<1>/NP{Z}){Y}){_}/NP{W}<2>){_} 2933 | 1 prep %f %l 2934 | 2 pobj %l %f 2935 | 2936 | # In-situ object wh-determiner 2937 | (S[wq]\(S[dcl]/NP))/N 2938 | 2 ((S[wq]{_}\(S[dcl]{Y}<1>/NP{Z*}){Y}){_}/N{Z}<2>){_} 2939 | 1 ignore 2940 | 2 det %f %l 2941 | 2942 | # Subject "how much" 2943 | (S[wq]/(S[dcl]\NP))/NP 2944 | 2 ((S[wq]{_}/(S[dcl]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 2945 | 1 ignore 2946 | 2 advmod %f %l 2947 | 2948 | ((NP\NP)/(S[pss]\NP))/NP 2949 | 3 (((NP{Y}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{W}){Z}){_}/NP{V}<3>){_} 2950 | 1 mod %f %l 2951 | 2 pobj %l %f 2952 | 3 acomp %l %f 2953 | 2954 | ((S[pss]\NP)/PP)/NP 2955 | 3 (((S[pss]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 2956 | 1 nsubjpass %l %f 2957 | 2 iobj %l %f =PP/NP 2958 | 2 acomp %l %f =PP/(S[ng]\NP) 2959 | 2 acomp %l %f =PP/(S[adj]\NP) 2960 | 3 dobj %l %f 2961 | 2962 | (((S[q]/NP)/PP)/(((S[b]\NP)/PP)/NP))/NP 2963 | 4 ((((S[q]{_}/NP{Y}<1>){_}/PP{Z}<2>){_}/(((S[b]{W}<3>\NP{V*}){W}/PP{Z*}){W}/NP{Y*}){W}){W}/NP{V}<4>){_} 2964 | 1 ignore 2965 | 2 ignore 2966 | 3 ignore 2967 | 4 ignore 2968 | # there is only one example of this category in the question data, and it looks wrong 2969 | 2970 | ((S[q]/PP)/NP)/NP 2971 | 3 (((S[q]{_}/PP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 2972 | 1 nsubj %l %f 2973 | 2 acomp %l %f 2974 | 3 iobj %l %f 2975 | 2976 | ((S[q]/PP)/PP)/NP 2977 | 3 (((S[q]{_}/PP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 2978 | 1 mod %f %l # don't have two iobj's 2979 | 2 iobj %l %f =PP/NP 2980 | 3 nsubj %l %f 2981 | 2982 | (((S[wq]/NP)/((S[q]/NP)/NP))/N)/(NP/N) 2983 | 4 ((((S[wq]{_}/NP{Y}<1>){_}/((S[q]{Z}<2>/NP{Y*}){Z}/NP{W*}){Z}){_}/N{W}<3>){_}/(NP{V}/N{V}){U}<4>){_} 2984 | 1 ignore 2985 | 2 ignore 2986 | 3 ignore 2987 | 4 ignore 2988 | # there is only one example of this category in the question data, and it looks like it has too many NPs 2989 | 2990 | (((S[wq]/PP)/((S[q]/PP)/NP))/N)/(NP/N) 2991 | 4 ((((S[wq]{_}/PP{Y}<1>){_}/((S[q]{Z}<2>/PP{Y*}){Z}/NP{W*}){Z}){_}/N{W}<3>){_}/(NP{V}/N{V}){U}<4>){_} 2992 | 1 ignore 2993 | 2 ignore 2994 | 3 ignore 2995 | 4 ignore 2996 | # this looks like it has too many NPs also 2997 | 2998 | # how in "how far is it from" 2999 | ((S[wq]/PP)/((S[q]/PP)/(S[adj]\NP)))/(S[adj]\NP) 3000 | 3 (((S[wq]{_}/PP{Y}<1>){_}/((S[q]{Z}<2>/PP{Y*}){Z}/(S[adj]{W*}\NP{V}){W*}){Z}){_}/(S[adj]{W}<3>\NP{V}){W}){_} 3001 | 1 ignore 3002 | 2 ignore 3003 | 3 advmod %f %l 3004 | 3005 | # adverb modifying S[q]/NP 3006 | (S/NP)\(S/NP) 3007 | 1 ((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_} 3008 | 1 advmod %f %l 3009 | 3010 | (S[wq]\(S[dcl]/PP))/(S[wq]\(S[dcl]/NP)) 3011 | 2 ((S[wq]{Y}\(S[dcl]{Z}<1>/PP{W}){Z}){Y}/(S[wq]{Y}<2>\(S[dcl]{V}/NP{U}){V}){Y}){_} 3012 | 1 iobj %f %l 3013 | 2 ignore 3014 | # see "In_II which_DDQ abbey_NN1 did_VDD he_PPHS1 see_VV0 her_PPHO1 ?_?" in the grammar dev files 3015 | # to get the same deps we'd need to relate "in" and "abbey" 3016 | # look into this at some point 3017 | 3018 | (((S[wq]/(S[pss]\NP))/((S[q]/(S[pss]\NP))/NP))/N)/(NP/N) 3019 | 4 ((((S[wq]{_}/(S[pss]{Y}<1>\NP{Z}){Y}){_}/((S[q]{W}<2>/(S[pss]{Y*}\NP{Z}){Y}){W}/NP{V*}){W}){_}/N{V}<3>){_}/(NP{U}/N{U}){T}<4>){_} 3020 | 1 ignore 3021 | 2 ignore 3022 | 3 ignore 3023 | 4 ignore 3024 | # only one example of this in the question data; looks like there are too many NPs 3025 | 3026 | (S[wq]/(S[q]/NP))/(S[wq]/(S[q]/NP)) 3027 | 1 ((S[wq]{Y}/(S[q]{Z}/NP{W}){Z}){Y}/(S[wq]{Y}<1>/(S[q]{Z}/NP{W}){Z}){W}){_} 3028 | 1 ignore 3029 | # only one example of this in the question data; and difficult to get looking like B+C 3030 | 3031 | (S[wq]/(S[q]/PP))/N 3032 | 2 ((S[wq]{_}/(S[q]{Y}<1>/PP{Z}){Y}){_}/N{W}<2>){_} 3033 | 1 ignore 3034 | 2 det %f %l 3035 | 3036 | ((S[wq]/(S[q]/PP))/N)/(NP/N) 3037 | 3 (((S[wq]{_}/(S[q]{Y}<1>/PP{Z}){Y}){_}/N{W}<2>){_}/(NP{V}/N{V}){U}<3>){_} 3038 | 1 mod %f %l 3039 | 2 dobj %l %f # not sure 3040 | 3 mod %2 %f 3041 | # based on "How many abbots can the abbey afford ?" from the RASP grammar development files 3042 | 3043 | # Adverbial 3044 | S[wq]\S[wq] 3045 | 1 (S[wq]{Y}\S[wq]{Y}<1>){_} 3046 | 1 advmod %f %l 3047 | 3048 | (S[wq]\S[wq])/(S[dcl]\NP) 3049 | 2 ((S[wq]{Y}\S[wq]{Y}<1>){_}/(S[dcl]{Z}<2>\NP{W}){Z}){_} 3050 | 1 mod %f %l 3051 | 2 acomp %l %f 3052 | 3053 | # Initial "when/where" when it is an argument e.g. When is Ford's birthday? 3054 | # (Now revised to NP not PP) 3055 | S[wq]/(S[q]/PP) 3056 | 1 (S[wq]{_}/(S[q]{Y}<1>/PP{Z}){Y}){_} 3057 | 1 ignore 3058 | 3059 | # fronted preposition with adjective question, adjunct: "for how long is an elephant pregnant?" 3060 | (S[wq]/S[q])/(S[wq]/(S[q]/(S[adj]\NP))) 3061 | 3 ((S[wq]{_}/S[q]{Y}<1>){_}/(S[wq]{W}/(S[q]{V}/(S[adj]{U}<3>\NP{T}){U}){V}){W}<2>){_} 3062 | 1 prep %l %f 3063 | 2 pobj %l %f 3064 | 3 advmod %1 %f 3065 | 3066 | (S[wq]/(S[q]/PP))/(S[wq]/(S[q]/(S[adj]\NP))) 3067 | 2 ((S[wq]{_}/(S[q]{Y}/PP{Z}){Y}<1>){_}/(S[wq]{W}/(S[q]{V}/(S[adj]{U}\NP{T}){U}){V}){W}<2>){_} 3068 | 1 mod %f %l 3069 | 2 mod %f %l 3070 | # not sure about this one 3071 | 3072 | # Inverted aux e.g. "how far is Pluto from" 3073 | ((S[q]/PP)/(S[adj]\NP))/NP 3074 | 3 (((S[q]{_}/PP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_}/NP{V}<3>){_} 3075 | 1 prep %l %f 3076 | 2 cop %f %l 3077 | 3 nsubjfix %l %f # Should be nsubj %2 %f but not being produced; post-process 3078 | 3079 | # Inverted aux e.g. "how far is it from" 3080 | ((S[q]/PP)/(S[adj]\NP))/NP[expl] 3081 | 3 (((S[q]{_}/PP{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_}/NP[expl]{V}<3>){_} 3082 | 1 prep %l %f 3083 | 2 cop %f %l 3084 | 3 nsubjfix %l %f # Should be nsubj %2 %f but not being produced; post-process 3085 | 3086 | # Adjectival argument question, e.g. "how far/big/high is ... ?" 3087 | (S[wq]/(S[q]/(S[adj]\NP)))/(S[adj]\NP) 3088 | 2 ((S[wq]{_}/(S[q]{Y}<1>/(S[adj]{Z}\NP{W}){Z}){V}){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 3089 | 1 ignore # the copula relation is done on the aux itself 3090 | 2 advmod %f %l 3091 | 3092 | # Adjectival adjunct question, e.g. "how far/fast does ... ?" 3093 | (S[wq]/S[q])/(S[adj]\NP) 3094 | 2 ((S[wq]{_}/S[q]{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 3095 | 1 advmod %f %2 3096 | 2 advmod %f %l 3097 | 3098 | S[wq]\(S[dcl]/NP) 3099 | 1 (S[wq]{_}\(S[dcl]{Y}<1>/NP{Z}){Y}){_} 3100 | ! (S[wq]{_}\(S[dcl]{Y}<1>/NP{_}){Y}){_} 3101 | 1 ignore 3102 | 3103 | (S[wq]\(S[dcl]/PP))/N 3104 | 2 ((S[wq]{_}\(S[dcl]{Y}<1>/PP{Z}){Y}){_}/N{W}<2>){_} 3105 | 1 ignore 3106 | 2 ignore 3107 | # is this category in the question data? 3108 | 3109 | (S[wq]\S[wq])/S[dcl] 3110 | 2 ((S[wq]{Y}\S[wq]{Y}<1>){_}/S[dcl]{Z}<2>){_} 3111 | 1 mod %f %l 3112 | 2 ccomp %l %f 3113 | 3114 | (((S[q]/PP)/PP)/(S[adj]\NP))/NP 3115 | 4 ((((S[q]{_}/PP{Y}<1>){_}/PP{Z}<2>){_}/(S[adj]{W}<3>\NP{V}){W}){_}/NP{U}<4>){_} 3116 | 1 ignore 3117 | 2 ignore 3118 | 3 ignore 3119 | 4 ignore 3120 | # is this category in the question data? 3121 | 3122 | ((S[pss]\NP)/NP)/PP 3123 | 3 (((S[pss]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/PP{W}<3>){_} 3124 | 1 nsubjpass %l %f 3125 | 2 dobj %l %f 3126 | 3 iobj %l %f =PP/NP 3127 | 3128 | NP/N 3129 | 1 (NP{Y}/N{Y}<1>){_} 3130 | 1 dobj %l %f # no exx in ccgbank. expect this is det? 3131 | 3132 | # Pied-piped preposition e.g. "by what nickname was so-and-so known" 3133 | (S[wq]/(S[q]/PP))/(S[wq]/(S[q]/NP)) 3134 | 2 ((S[wq]{_}/(S[q]{Y}<1>/PP{Z}){Y}){_}/(S[wq]{W}/(S[q]{V}/NP{U}){V}){W}<2>){_} 3135 | 1 prep %f %l 3136 | 2 pobj %l %f 3137 | 3138 | # Fronted adverbial adjunct wh-PP e.g. "during what war did such-and-such happen" 3139 | (S[wq]/S[q])/(S[wq]/(S[q]/NP)) 3140 | 2 ((S[wq]{_}/S[q]{Y}<1>){_}/(S[wq]{W}/(S[q]{V}/NP{U}){V}){W}<2>){_} 3141 | 1 prep %f %l 3142 | 2 pobj %l %f 3143 | 3144 | # Adverbial adjunct non-fronted wh-PP e.g. "such-and-such happened during which war" 3145 | (S[wq]\S[dcl])/(S[wq]\(S[dcl]/NP)) 3146 | 2 ((S[wq]{_}\S[dcl]{Y}<1>){_}/(S[wq]{W}\(S[dcl]{V}/NP{U}){V}){W}<2>){_} 3147 | 1 prep %f %l 3148 | 2 pobj %l %f 3149 | 3150 | ((S[wq]/(S[q]/PP))\(S[wq]/(S[q]/PP)))/NP 3151 | 2 (((S[wq]{Y}/(S[q]{Z}/PP{W}){Y}){_}\(S[wq]{Y}<1>/(S[q]{Z}/PP{W}){Z}){Y}){_}/NP{V}<2>){_} 3152 | 1 mod %f %l 3153 | 2 dobj %l %f # not sure 3154 | 3155 | # For "how many years was he with GE" 3156 | ((S[wq]/S[q])/N)/(NP/N) 3157 | 3 (((S[wq]{_}/S[q]{U}<1>){_}/N{Z}<2>)/(NP/N){Y}<3>){_} 3158 | 1 ignore 3159 | 2 tmod %1 %f 3160 | 3 advmod %f %l 3161 | 3162 | # PP modifier e.g. "where on the body is ..." 3163 | ((S[wq]/(S[q]/NP))\(S[wq]/(S[q]/NP)))/NP 3164 | 2 (((S[wq]{Y}/(S[q]{Z}/NP{W}){Y}){_}\(S[wq]{Y}<1>/(S[q]{Z}/NP{W}){Z}){Y}){_}/NP{V}<2>){_} 3165 | 1 prep %f %l 3166 | 2 pobj %l %f 3167 | 3168 | # "How" in "how many/much" 3169 | ((S[wq]/(S[q]/NP))/N)/(NP/N) 3170 | 3 (((S[wq]{_}/(S[q]{Y}<1>/NP{Z*}){Y}){_}/N{Z}<2>){_}/(NP{W}/N{W}){V}<3>){_} 3171 | 1 ignore 3172 | 2 amod %f %3 # Should this be det rather than amod? Are we doing nn/amod? 3173 | 3 advmod %f %l # (following de Marneffe p.c.) 3174 | 3175 | ((S[wq]/PP)/N)/(NP/N) 3176 | 3 (((S[wq]{_}/PP{Y}<1>){_}/N{Z}<2>){_}/(NP{W}/N{W}){V}<3>){_} 3177 | 1 mod %f %l 3178 | 2 dobj %l %f # not sure 3179 | 3 mod %2 %f 3180 | 3181 | (S[wq]\(S[dcl]/PP))/(S[wq]\(S[dcl]/PP)) 3182 | 1 ((S[wq]{Y}\(S[dcl]{Z}/PP{W}){Z}){Y}/(S[wq]{Y}<1>\(S[dcl]{Z}/PP{W}){Z}){Y}){_} 3183 | 1 mod %f %l 3184 | # is this category in the question data? 3185 | 3186 | LQU 3187 | 1 LQU{_} 3188 | 3189 | RQU 3190 | 1 RQU{_} 3191 | 3192 | (((S[b]\NP)/NP)/PP)/NP 3193 | 4 ((((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/PP{W}<3>){_}/NP{V}<4>){_} 3194 | 1 nsubj %l %f 3195 | 2 iobj %l %f 3196 | 3 iobj %l %f 3197 | 4 dobj %l %f 3198 | # is this category in the question data? 3199 | 3200 | ((S[pss]\NP)/NP)/NP 3201 | 3 (((S[pss]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 3202 | 1 nsubjpass %l %f 3203 | 2 iobj %l %f 3204 | 3 dobj %l %f 3205 | # is this category in the question data? 3206 | 3207 | (N/N)\N 3208 | 2 ((N{Y}/N{Y}<1>){_}\N{Z}<2>){_} 3209 | 1 ignore 3210 | 2 ignore 3211 | # is this category in the question data? 3212 | 3213 | # Inverted aux 3214 | (S[q]/(S[pss]\NP))/NP 3215 | 2 ((S[q]{_}/(S[pss]{Y}<1>\NP{Z*}){Y}){_}/NP{Z}<2>){_} 3216 | ! ((S[q]{Y}/(S[pss]{Y}<1>\NP{Z*}){Y}){Y}/NP{Z}<2>){_} 3217 | 1 auxpass %f %l 3218 | 2 ignore 3219 | 3220 | # Inverted aux 3221 | (S[q]/PP)/NP 3222 | 2 ((S[q]{_}/PP{Y}<1>){_}/NP{Z}<2>){_} 3223 | 1 prep %l %f 3224 | 2 nsubj %l %f 3225 | 3226 | #Subject wh-determiner 3227 | (S[wq]/(S[dcl]\NP))/N 3228 | 2 ((S[wq]{Z}/(S[dcl]{Y}<1>\NP{Z*}){Y}){Z}/N{Z}<2>){_} 3229 | 1 ignore 3230 | 2 det %f %l 3231 | 3232 | NP/PP 3233 | 1 (NP{_}/PP{Y}<1>){_} 3234 | 1 iobj %l %f 3235 | 3236 | ((S[q]/NP)/NP)/NP 3237 | 3 (((S[q]{_}/NP{Y}<1>){_}/NP{Z}<2>){_}/NP{W}<3>){_} 3238 | 1 ignore 3239 | 2 ignore 3240 | 3 ignore 3241 | # is this category in the question data? 3242 | 3243 | # extra categories added to the end for Matt's CCG-PropBank work 3244 | (((S[dcl]\NP)/PP)/PP)/PP 3245 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_}/PP{V}<4>){_} 3246 | 1 mod %f %l 3247 | 2 iobj %l %f 3248 | 3 iobj %l %f 3249 | 4 iobj %l %f 3250 | 3251 | PP/N[num] 3252 | 1 (PP{_}/N[num]{Y}<1>){_} 3253 | 1 pobj %l %f 3254 | 3255 | PP/N 3256 | 1 (PP{_}/N{Y}<1>){_} 3257 | 1 pobj %l %f 3258 | 3259 | ((S[pt]\NP)/PP)/PP 3260 | 3 (((S[pt]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_} 3261 | 1 mod %f %l 3262 | 2 iobj %l %f 3263 | 3 iobj %l %f 3264 | 3265 | ((PP/PP)/(PP/PP))\(S[adj]\NP) 3266 | 2 (((PP{_}/PP{_}){_}/(PP{_}/PP{_}){_}<1>)\(S[adj]{Y}\NP{Z}){Y}<2>){_} 3267 | 1 mod %f %l 3268 | 2 mod %f %l 3269 | 3270 | ((S[pt]\NP)/NP)/PP 3271 | 3 (((S[pt]{_}\NP{Y}<1>){_}/NP{Z}<2>){_}/PP{W}<3>){_} 3272 | 1 mod %f %l 3273 | 2 dobj %l %f 3274 | 3 iobj %l %f 3275 | 3276 | (((S[pt]\NP)/PP)/PP)/PP 3277 | 4 ((((S[pt]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_}/PP{V}<4>){_} 3278 | 1 mod %f %l 3279 | 2 iobj %l %f 3280 | 3 iobj %l %f 3281 | 4 iobj %l %f 3282 | 3283 | PP/(S[dcl]\NP) 3284 | 1 (PP{_}/(S[dcl]{Y}\NP{Z}){Y}<1>){_} 3285 | 1 advcl %l %f 3286 | 3287 | (PP\PP)/((S\NP)\(S\NP)) 3288 | 1 ((PP{Y}\PP{Y}<1>){Y}/((S{Z}\NP{V}){Z}\(S{Z}\NP{V}){Z}){Z}<2>){_} 3289 | 1 mod %f %l 3290 | 2 mod %f %l 3291 | 3292 | (((S[pt]\NP)/PP)/PP)/NP 3293 | 4 ((((S[pt]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/PP{W}<3>){_}/NP{V}<4>){_} 3294 | 1 mod %f %l 3295 | 2 iobj %l %f 3296 | 3 iobj %l %f 3297 | 4 dobj %l %f 3298 | 3299 | (((S[dcl]\NP)/PP)/NP)/PP 3300 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_}/PP{V}<4>){_} 3301 | 1 nsubj %l %f 3302 | 2 iobj %l %f =PP/NP 3303 | 2 acomp %l %f =PP/(S[ng]\NP) 3304 | 2 acomp %l %f =PP/(S[adj]\NP) 3305 | 3 dobj %l %f 3306 | 4 iobj %l %f =PP/NP 3307 | 3308 | ((S[dcl]\NP)/(S[ng]\NP))/PP 3309 | 3 (((S[dcl]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_}/PP{V}<3>){_} 3310 | 1 ignore =aux 3311 | 1 nsubj %l %f 3312 | 2 aux %f %l =aux 3313 | 2 acomp %l %f 3314 | 3 iobj %l %f 3315 | 3316 | (((S[b]\NP)/PP)/NP)/PP 3317 | 4 ((((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_}/PP{V}<4>){_} 3318 | 1 nsubj %l %f 3319 | 2 iobj %l %f 3320 | 3 dobj %l %f 3321 | 4 iobj %l %f 3322 | 3323 | (((S[dcl]\NP)/(S[to]\NP))/PP)/PP 3324 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{V}<3>){_}/PP{U}<4>){_} 3325 | 1 nsubj %l %f 3326 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 3327 | 3 iobj %l %f =PP/NP 3328 | 4 iobj %l %f =PP/NP 3329 | 3330 | (((S[ng]\NP)/PP)/NP)/PP 3331 | 4 ((((S[ng]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_}/PP{V}<4>){_} 3332 | 1 nsubj %l %f 3333 | 2 iobj %l %f =PP/NP 3334 | 2 acomp %l %f =PP/(S[ng]\NP) 3335 | 2 acomp %l %f =PP/(S[adj]\NP) 3336 | 3 dobj %l %f 3337 | 4 iobj %l %f 3338 | 3339 | ((S[b]\NP)/(S[ng]\NP))/PP 3340 | 3 (((S[b]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{W*}){Z}){_}/PP{V}<3>){_} 3341 | 1 ignore =aux 3342 | 1 nsubj %l %f 3343 | 2 aux %f %l =aux 3344 | 2 acomp %l %f 3345 | 3 iobj %l %f 3346 | 3347 | NP/N[num] 3348 | 1 (NP{_}/N[num]{Y}<1>){_} 3349 | 1 num %f %l # $ is head 3350 | 3351 | (((S[dcl]\NP)/(S[to]\NP))/PP)/NP 3352 | 4 ((((S[dcl]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{W}){Z}){_}/PP{V}<3>){_}/NP{U}<4>){_} 3353 | 1 nsubj %l %f 3354 | 2 xcomp %l %k =(S[to]\NP)/(S[b]\NP) 3355 | 3 iobj %l %f =PP/NP 3356 | 4 dobj %l %f 3357 | 3358 | # New bio cats 3359 | (N\N)\(N\N) 3360 | 1 ((N{Y}\N{Y}){Z}\(N{Y}\N{Y}){Z}<1>){_} 3361 | 1 advmod %f %l 3362 | 3363 | ((S[pss]\NP)/PP)/(PP/NP) 3364 | 3 (((S[pss]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/(PP{W}/NP{X})<3>){_} 3365 | 1 nsubjpass %l %f 3366 | 2 prep %l %f 3367 | 3 prep %l %f 3368 | 3369 | # This is in ccgbank, presumably omitted from markedup file in error 3370 | ((S[adj]\NP)/(S[adj]\NP))/NP 3371 | 2 (((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}<1>\NP{Z}){Y}){_}/NP{W}<2>){_} 3372 | 1 advmod %f %2 3373 | 2 det %f %l =fulldet 3374 | 2 pobj %l %f 3375 | 3376 | # Ditto 3377 | ((S[adj]\NP)/(S[adj]\NP))\((S[adj]\NP)/(S[adj]\NP)) 3378 | 1 (((S[adj]{Y}\NP{Z}){Y}\(S[adj]{Y}\NP{Z}){Y}){W}/((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}\NP{Z}){Y}){W}<1>){_} 3379 | 1 amod %f %l 3380 | 3381 | # Ditto 3382 | # This needs some work 3383 | (NP\NP)/S[em] 3384 | 2 ((NP{Y}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 3385 | ! ((NP{Y}\NP{Y}<1>){_}/S[em]{Z}<2>){_} 3386 | 1 ignore 3387 | 2 rcmod %1 %c 3388 | 3389 | (S/NP)/(S/NP) 3390 | 1 ((S[X]{Y}/NP{Z}){Y}/(S[X]{Y}<1>/NP{Z}){Y}){_} 3391 | 1 advmod %f %l 3392 | 3393 | (((S\NP)\(S\NP))/(NP\NP))/NP 3394 | 3 ((((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_}/(NP\NP){W}<2>)/NP{V})<3> 3395 | 1 prep %f %l 3396 | 2 mod %f %l 3397 | 3 pobj %l %f 3398 | 3399 | (((N/N)/(N/N))\((N/N)/(N/N)))/N 3400 | 2 ((((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){W}/((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}){W}<2>)/N{V}<1>){_} 3401 | 1 pobj %l %f 3402 | 2 prep %f %l 3403 | 3404 | # Ditto 3405 | (S[dcl]\NP[expl])/(S[adj]\NP) 3406 | 2 ((S[dcl]{_}\NP[expl]{Y}<1>){_}/(S[adj]{Z}<2>\NP{W}){Z}){_} 3407 | ! ((S[dcl]{Z}\NP[expl]{Y}<1>){Z}/(S[adj]{Z}<2>\NP{Y}){Z}){_} 3408 | 1 ignore =aux 3409 | 2 cop %f %l 3410 | 3411 | ((N/N)/(N/N))\NP 3412 | 2 (((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}<2>){_}/NP{W}<1>){_} 3413 | 1 poss %f %2 3414 | 2 possessive %f %l 3415 | 3416 | ((S[adj]\NP)/(S[adj]\NP))\NP 3417 | 2 (((S[adj]{Y}\NP{Z}){Y}/(S[adj]{Y}<1>\NP{Z}){Y}){_}\NP{W}<2>){_} 3418 | 1 prep %f %l 3419 | 2 pobj %l %f 3420 | 3421 | ((S/S)/(S/S))/(S[asup]\NP) 3422 | 2 (((N{Y}/N{Y}){Z}/(N{Y}/N{Y}){Z}<2>){_}/(S[asup]{W}<1>\NP{V}){W}){_} 3423 | 1 xcomp %l %f 3424 | 2 mod %f %l 3425 | 3426 | ((N/N)\NP)/NP 3427 | 2 (((N{Y}/N{Y}){Z}\NP{Z}<1>){_}/NP{W}<2>){_} 3428 | 1 prep %f %l 3429 | 2 pobj %l %f 3430 | 3431 | ((NP\NP)/(S[to]\NP))\((NP\NP)/NP) 3432 | 3 (((NP{Y}\NP{Y}<1>){_}/(S[to]\NP){Z}<2>){_}\((NP{W}\NP{W}){V}/NP{U}){V}<3>){_} 3433 | 1 advcl %f %2 3434 | 2 mod %f %3 3435 | 3 dep %f %l # following Stanford parser 3436 | 3437 | (S[inv]/NP)/(S[pss]\NP) 3438 | 2 ((S[inv]{_}/NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 3439 | ! ((S[inv]{Z}/NP{Y}<1>){Z}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 3440 | 1 nsubj %l %f 3441 | 2 cop %f %l 3442 | 3443 | ((N/N)\(N/N))/PP 3444 | 2 (((N{Y}/N{Y}){Z}\(N{Y}/N{Y}){Z}<1>){_}/PP{W}<2>){_} 3445 | 1 prep %f %l 3446 | 2 dep %l %f 3447 | 3448 | PP/(N/N) 3449 | 1 (PP{_}/(N{Y}/N{Y})<1>){_} 3450 | 1 pobj %l %f 3451 | 3452 | (S[dcl]\(S[ng]\NP))/PP 3453 | 2 ((S[dcl]{_}\(S[ng]{Y}<1>\NP{Z}){Y}){_}/PP{W}<2>){_} 3454 | 1 acomp %l %f 3455 | 2 nsubjpass %l %f 3456 | 3457 | (N/N)\(N/N) 3458 | 1 ((N{Y}/N{Y}){Z}\(N{Y}/N{Y}){Z}<1>){_} 3459 | 1 advmod %l %f 3460 | 3461 | ((N/N)\(N/N))\((N/N)\(N/N)) 3462 | 2 (((N{Y}/N{Y}){_}\(N{Y}/N{Y})<1>){_}\((N{X}/N{X}){_}\(N{X}/N{X}))<2>){_} 3463 | 1 nmod %l %f 3464 | 2 nmod %l %f 3465 | 3466 | ((N/N)\(N/N))/S[dcl] 3467 | 2 (((N{Y}/N{Y})\(N{Y}/N{Y})<1>){_}/S[dcl]{Z}<2>){_} 3468 | 1 rcmod %f %2 3469 | 2 dep %f %l 3470 | -------------------------------------------------------------------------------- /downloads/models-1.02.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/models-1.02.tgz -------------------------------------------------------------------------------- /downloads/pos_bio-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/pos_bio-1.00.tgz -------------------------------------------------------------------------------- /downloads/ptb_chunk-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/ptb_chunk-1.00.tgz -------------------------------------------------------------------------------- /downloads/ptb_muc-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/ptb_muc-1.00.tgz -------------------------------------------------------------------------------- /downloads/ptb_pos-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/ptb_pos-1.00.tgz -------------------------------------------------------------------------------- /downloads/super_bio-1.00.tgz: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chbrown/candc/bd4fbe5a20797328a6623736acdb95111c581312/downloads/super_bio-1.00.tgz --------------------------------------------------------------------------------